WO2024255225A1 - 在线客服咨询排队时长的预测方法和装置 - Google Patents

在线客服咨询排队时长的预测方法和装置 Download PDF

Info

Publication number
WO2024255225A1
WO2024255225A1 PCT/CN2024/071118 CN2024071118W WO2024255225A1 WO 2024255225 A1 WO2024255225 A1 WO 2024255225A1 CN 2024071118 W CN2024071118 W CN 2024071118W WO 2024255225 A1 WO2024255225 A1 WO 2024255225A1
Authority
WO
WIPO (PCT)
Prior art keywords
customer service
user
queuing
queue
time
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2024/071118
Other languages
English (en)
French (fr)
Inventor
何峰
何刚
康志恒
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Jingdong Century Trading Co Ltd
Beijing Wodong Tianjun Information Technology Co Ltd
Original Assignee
Beijing Jingdong Century Trading Co Ltd
Beijing Wodong Tianjun Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Jingdong Century Trading Co Ltd, Beijing Wodong Tianjun Information Technology Co Ltd filed Critical Beijing Jingdong Century Trading Co Ltd
Publication of WO2024255225A1 publication Critical patent/WO2024255225A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3329—Natural language query formulation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/338—Presentation of query results
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/34—Browsing; Visualisation therefor
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/01—Customer relationship services
    • G06Q30/015—Providing customer assistance, e.g. assisting a customer within a business location or via helpdesk
    • G06Q30/016—After-sales

Definitions

  • the present disclosure relates to the field of computer technology, and in particular to a method and device for predicting the queuing time of online customer service consultation.
  • NLP Natural Language Processing
  • the inventors found that the existing technology, whether estimating the queuing time based on statistical methods or based on NLP technology, has a large estimation error of the queuing time and low prediction performance.
  • the embodiments of the present disclosure provide a method and device for predicting the queuing time for online customer service consultation, which greatly reduces the error of the prediction result and improves the prediction accuracy and prediction performance of the queuing time.
  • a method for predicting the queuing time of online customer service consultation comprising:
  • the customer service status characteristics include the number of currently available customer service staff and the number of sessions being processed by each customer service staff;
  • the queuing time estimation model is used to predict the queuing time of the user according to the queuing characteristics of the user, the number of currently available customer service representatives, and the number of sessions being processed by each customer service representative.
  • the queuing characteristics of the user include the queue number of the user in the waiting queue and the queue number of the user in the skill group; according to the queuing characteristics of the user, the currently available number of customer services and the number of sessions being processed by each customer service, the queuing time prediction model is used to predict the queuing time of the user, including: according to the queue number of the user in the waiting queue and the queue number of the user in the skill group, the currently available number of customer services and the number of sessions being processed by each customer service, the queuing time prediction model is used to predict the queuing time of the user.
  • the method further includes: determining a queuing time period corresponding to the user according to the sending time of the user's online customer service consultation request; using a queuing time estimation model to predict the queuing time of the user according to the queuing characteristics of the user, the current number of available customer service representatives, and the number of sessions being processed by each customer service representative, including: determining a queuing time period corresponding to the user, the queue number of the user in the waiting queue, and the number of sessions in the skill group in which the user is in the waiting queue, according to the ... sending time of the user's online customer service consultation request; The queue number, the number of currently available customer service representatives, and the number of sessions being processed by each customer service representative are used to predict the queue time of the user using a queue time estimation model.
  • the queuing time prediction model is constructed in the following manner: obtaining user online customer service consultation session data in a recent period; extracting session features corresponding to each session from the user online customer service consultation session data, the session features including the request for online consultation time, the queuing time period, the skill group to which one belongs, the queue number of the user in the waiting queue and the queue number of the user in the skill group to which one belongs, the number of currently available customer services in the skill group to which one belongs and the number of sessions being processed by each customer service, as well as the actual online consultation time; obtaining the queuing time corresponding to each session based on the request for online consultation time and the actual online consultation time corresponding to each session; for each session, using the queuing time corresponding to the session to mark the session features corresponding to the session to generate training data corresponding to the session; and performing model training based on the training data corresponding to each session to construct the queuing time prediction model.
  • the queuing time prediction model after using the queuing time prediction model to predict the queuing time of the user, it also includes: in response to the user's specified operation trigger event, re-acquiring the user's queuing characteristics and the customer service status characteristics of the skill group, the customer service status characteristics including the current number of available customer service staff and the number of sessions being processed by each customer service; based on the user's queuing characteristics, the current number of available customer service staff and the number of sessions being processed by each customer service, using the queuing time prediction model to predict the user's queuing time to update the user's queuing time.
  • the queuing time prediction model after using the queuing time prediction model to predict the queuing time of the user, it also includes: in response to detecting that a customer service has ended the session being processed and released resources, taking out the queued user from the waiting queue and assigning it to the customer service, refreshing the waiting queue and updating the customer service status; when the customer service status is still that all customer services are busy, re-acquiring the queuing characteristics of the user and the customer service status characteristics of the skill group, the customer service status characteristics including the current number of available customer services and the number of sessions being processed by each customer service; based on the queuing characteristics of the user, the current number of available customer services and the number of sessions being processed by each customer service, using the queuing time prediction model to predict the queuing time of the user, so as to update the queuing time of the user.
  • the method further includes: when the customer service status is that there is a customer service who is not busy, allocating the user to the customer service who is not busy.
  • a device for predicting the queuing time of online customer service consultation comprising:
  • a status acquisition module configured to determine the skill group corresponding to the user and acquire the customer service status of the skill group in response to receiving an online customer service consultation request from the user;
  • a queue adding module used for adding the user to the waiting queue when the customer service status is that all customer service staff are busy
  • a feature acquisition module used to acquire the queuing features of the user and the customer service status features of the skill group, wherein the customer service status features include the number of currently available customer service staff and the number of sessions being processed by each customer service staff;
  • the duration prediction module is used to predict the queuing time of the user using a queuing time estimation model according to the queuing characteristics of the user, the number of currently available customer service staff and the number of sessions being processed by each customer service staff.
  • an electronic device comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for predicting the waiting time for online customer service consultation provided in an embodiment of the present disclosure.
  • a computer-readable medium on which a computer program is stored.
  • the program is executed by a processor, the method for predicting the online customer service consultation queue length provided in an embodiment of the present disclosure is implemented.
  • One embodiment disclosed above has the following advantages or beneficial effects: by responding to receiving an online customer service consultation request from a user, determining the skill group corresponding to the user and obtaining the customer service status of the skill group; when the customer service status is that all customer service staff are busy, adding the user to the waiting queue; obtaining the user's queuing characteristics and the customer service status characteristics of the skill group, and obtaining the customer service status of the skill group;
  • the service status characteristics include the current number of available customer service staff and the number of sessions being handled by each customer service staff.
  • the technical solution of using the queue time estimation model to predict the user's queue time based on the user's queue characteristics, the current number of available customer service staff and the number of sessions being handled by each customer service staff can predict the current user's queue time based on the status of online customer service staff that can receive the user in the customer service system at the time the current user comes online, the status of the sessions being received by these customer service staff, and the user's queue characteristics, which greatly reduces the error of the prediction result and improves the prediction accuracy and prediction performance of the queue time.
  • FIG1 is a schematic diagram of the main steps of a method for predicting the queuing time of online customer service consultation according to an embodiment of the present disclosure
  • FIG2 is a schematic diagram of a process for predicting the length of time to queue for online customer service consultation according to an embodiment of the present disclosure
  • FIG3 is a schematic diagram of the error of the prediction result of the queuing time according to an embodiment of the present disclosure
  • FIG4 is a schematic diagram of main modules of a device for predicting the queuing time for online customer service consultation according to an embodiment of the present disclosure
  • FIG5 is an exemplary system architecture diagram in which embodiments of the present disclosure may be applied.
  • FIG. 6 is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present disclosure.
  • online customer service systems have become an important part of e-commerce websites.
  • online customer service human resources are limited, and customer service staff need to take into account both busy and idle times in a day.
  • customer service reception pressure is high, resulting in some users being unable to access manual customer service in time and entering the waiting queue.
  • the consultation queue is to reduce the peak and fill the valley of the number of consultation manual users, make full use of customer service resources, and allow users to access manual customer service after a short wait, rather than leaving a message without a customer service reply, to maximize the user pick-up rate and improve the user experience.
  • the core is to inform the user of the expected waiting time, which directly determines whether the user chooses to queue, and the accuracy of the estimated waiting time also directly affects the user's queuing experience. Therefore, the estimation of the waiting time is realized through an algorithm, based on the frequency of historical customer service ending the conversation, and the prediction of how long each ongoing conversation will end, as well as the different users queuing under different skill groups, to comprehensively calculate the user's queuing time.
  • self-operated merchants have tens of thousands of customer service staff who need to be received at the same time.
  • the relatively large customer service system has tens of thousands of customer service staff who need to be managed and received at the same time.
  • the statistical method is to count the number of people in the queue and multiply the average time of queuing by the number of people to get the estimated waiting time.
  • the method based on NLP technology to predict the end time of a single-call session is to predict the end time of a single-call session and accumulate the estimated end time of all ongoing sessions of all customer service staff who can receive the user to get the waiting time for a single user.
  • the final estimated queuing time requires the accumulation of the predicted time of multiple conversations, and the error will be infinitely magnified; (2) It is necessary to predict the estimated end time of all conversations currently being received by the online customer service.
  • the prediction of a user's queuing time requires thousands of predictions, which has low performance and low feasibility.
  • the present disclosure provides a method for predicting the queuing time for online customer service consultation, and its implementation principle is: according to the current user entry time, the online customer service status that can receive the user in the current customer service system, as well as the session status that these customer service are receiving, and the position of the user in the waiting queue are obtained, and relevant features are extracted.
  • the actual queuing time of historical users queuing under the current features is used as a label, and model learning is performed based on the machine learning regression method to predict the estimated queuing time when the current user enters the waiting queue, and the queuing time is divided into multiple slices, which are returned to the front end to be presented to the user.
  • Fig. 1 is a schematic diagram of the main steps of the method for predicting the online customer service consultation queue length according to an embodiment of the present disclosure. As shown in Fig. 1, the method for predicting the online customer service consultation queue length according to an embodiment of the present disclosure mainly includes the following steps S101 to S104.
  • Step S101 In response to receiving a user's online customer service consultation request, determine the user When a user requests online customer service, the system will first determine the skill group corresponding to the online customer service request based on the business scope. Each skill group corresponds to an online customer service task for a certain business.
  • Each skill group may include multiple customer service representatives, and each customer service representative may also correspond to multiple skill groups, and each customer service representative can handle multiple online customer service consultation sessions at the same time.
  • the customer service status of the skill group is that all customers are busy. Otherwise, if there are customer service representatives who have not handled sessions, or if the number of sessions currently handled by a customer service representative does not reach the limit of the number of sessions that the customer service representative can handle, the customer service status of the skill group is considered to be that there are customers available.
  • Step S102 When the customer service status is that all customer services are busy, the user is added to the waiting queue.
  • all customer services of the skill group are processing sessions, and the number of sessions currently processed by each customer service is the limit of the number of sessions that the customer service can process, all customer services are busy, and at this time, the user is added to the waiting queue.
  • Step S103 Obtain the queuing characteristics of the user and the customer service status characteristics of the skill group, wherein the customer service status characteristics include the number of currently available customer service staff and the number of sessions being processed by each customer service staff.
  • the queuing time of the user will be estimated based on the user's queuing status and the status characteristics of the customer service of the skill group.
  • the queuing characteristics of the user mainly include the queue number of the user in the waiting queue and the queue number of the user in the skill group.
  • the waiting queue stores all the queuing users who need online customer service consultation and their corresponding online customer service consultation requests, and also stores the skill group information corresponding to each user. Therefore, the first queue number of the user in the waiting queue and the second queue number of the user in the corresponding skill group can be obtained according to the information of the waiting queue.
  • Step S104 predicting the queuing time of the user using a queuing time estimation model according to the queuing characteristics of the user, the number of currently available customer service representatives, and the number of sessions being processed by each customer service representative.
  • the queuing characteristics of the user include the queue number of the user in the waiting queue and the queue number of the user in the skill group. Furthermore, according to the queuing characteristics of the user, the number of currently available customer service representatives, and the number of sessions being processed by each customer service representative, the queuing duration prediction model is used to predict the queuing duration of the user, which may specifically include: according to the queue number of the user in the waiting queue and the queue number of the user in the skill group, the number of currently available customer service representatives, and the number of sessions being processed by each customer service representative, the queuing duration prediction model is used to predict the queuing duration of the user.
  • a queue time estimation model before using a queue time estimation model to predict the queue time of the user according to the queue characteristics of the user, the current number of available customer service staff, and the number of sessions being processed by each customer service staff, it may also include: determining the queue time period corresponding to the user according to the sending time of the user's online customer service consultation request.
  • using a queue time estimation model to predict the queue time of the user may specifically include: using a queue time estimation model to predict the queue time of the user according to the queue time period corresponding to the user, the queue number of the user in the waiting queue and the queue number of the user in the skill group, the current number of available customer service staff, and the number of sessions being processed by each customer service staff.
  • the queuing time estimation model is constructed in the following manner: obtaining user online customer service consultation session data in a recent period; extracting session features corresponding to each session from the user online customer service consultation session data, the session features including the time of requesting online consultation, the queuing period, the skill group to which they belong, the queue number of the user in the waiting queue and the queue number of the user in the skill group to which they belong, the number of currently available customer services of the skill group to which they belong and the number of sessions being processed by each customer service, as well as the actual online consultation time; obtaining the queuing time corresponding to each session according to the time of requesting online consultation and the actual online consultation time corresponding to each session; for each session, marking the session features corresponding to the session using the queuing time corresponding to the session to generate training data corresponding to the session; and performing model training based on the training data corresponding to each session to construct the queuing time prediction model.
  • the present disclosure when constructing a queue time estimation model, the present disclosure first obtains the user online customer service consultation session data in the recent period, for example, the user queue session data, customer service reception data, and the actual queue time data in the past month can be screened.
  • the user queue session data is used to extract model training features and labels
  • the customer service reception data is used to calculate the user's online time, the number of customer service staff who can currently receive the user, and the number of ongoing sessions with these customer service staff
  • the actual queue time is obtained by burying the user's queue time to obtain the time between the user starting to queue and the conversation with the manual customer service.
  • the embodiment of the present disclosure needs to construct user portrait features and customer service portrait features respectively. At the same time, these features must be available at any time during the user queue process.
  • the feature definition is as follows:
  • Queuing time period divided into hours, the time period when users come in within the hour is numbered, because the time for users to queue up to access the line is different in different time periods. For example, if you queue up in the early morning, the number of customer service staff available is small, and the queuing time is usually more than 10 minutes. Therefore, the queuing time period can affect the actual queuing time of users;
  • the user's queue number in the waiting queue Statistics on the application customer service system, the sequence number of the current user in the waiting queue. The earlier the queue number, the shorter the waiting time;
  • the user's queue number in the skill group to which they belong Since users first navigate to the corresponding skill group through the system and queue under the skill group before they can access the line, the queue number under the skill group needs to be recorded. If the overall queue number is large, but the number of queues under the skill group is small, it means that the current incoming business is unevenly distributed. Users can still access the line in a shorter queue time under the skill group with fewer queues.
  • the number of currently available agents in the skill group that is, the total number of agents in the customer service system who have added the user's incoming skill group. The more agents there are, the faster the ongoing sessions are released and the shorter the user's waiting time.
  • Number of sessions being processed by customer service The number of sessions that customer service that can receive the current queued users is currently receiving at the same time. The more sessions there are, the greater the probability that the session will be released soon. The shorter the queue time;
  • Skill group identifier When a user comes in, he or she is first navigated to the corresponding skill group through the system. Due to business differences and the number of customer service personnel, the queuing time for different skill groups is also different. Therefore, this feature is also an important feature for predicting the user's queuing time.
  • the embodiment of the present disclosure needs to construct a label of the user's actual queue time for model learning.
  • the user can be buried at the time when the user starts queuing and ends queuing, and the user's actual queue time can be recorded, which can be used as a label for model training data.
  • the point is not buried, the user's actual queue time cannot be obtained, so it is necessary to construct a method to simulate the user's actual queue entry time, so as to use it as a label for model learning.
  • the specific method is as follows: with minutes as the granularity, take the conversation that generates messages within each minute, assume that these message users enter the line at the beginning of one minute to generate a queue at the same time, and then take the conversation that is happening at this moment, obtain the time from the actual end time of each conversation to the current moment, sort them from small to large according to the time length, and traverse the queue users and the ongoing conversations in turn. If the customer service received by the ongoing conversation can receive the queue user skill group, then the end time of this conversation is the waiting time of the current queue user, and so on. Theoretically, if the customer service has not grabbed the order, is not online or is suspended, and the upper limit of the connection mode has not been adjusted, then the queued user may be connected only after the current ongoing session ends.
  • Table 1 shows the queue information of queuing users according to an embodiment of the present disclosure.
  • Table 2 shows the currently ongoing conversation queue, which mainly includes the customer service number (including the skill group information mounted by the customer service), the actual end time of the conversation (in seconds), and the matching queue user index.
  • customer service representative A3-1005 in the online customer service can receive this skill group, so after the customer service representative's most recent conversation ends, the user in the queue can get on the line, so the actual end time of the conversation is 5 seconds, which is the waiting time for the user in the queue.
  • the queuing time corresponding to each session can be obtained.
  • the queuing time corresponding to the session is used to mark the session features corresponding to the session to generate the training data corresponding to the session.
  • model training is performed based on the training data corresponding to each session to construct a queuing time estimation model.
  • a GBDT Gradient Boosting Decision Tree
  • the user's queue time can be predicted based on the user's corresponding queue period, the user's queue characteristics, the current number of available customer service representatives, and the number of sessions being handled by each customer service representative.
  • the queue time estimation model after using the queue time estimation model to predict the queue time of the user, it may also include: in response to the specified operation triggering event of the user, re-acquiring the queue characteristics of the user and the customer service status characteristics of the skill group, the customer service status characteristics including the current number of available customer service staff and the number of sessions being handled by each customer service; based on the queue characteristics of the user, the current number of available customer service staff and the number of sessions being handled by each customer service, using the queue time estimation model to predict the queue time of the user to update the queue time of the user.
  • the specified operation is, for example, click, slide, return, switch screen and other operations.
  • the user When the user generates event operations such as click, slide, return, switch screen, etc., it will trigger the prediction of the user's queue time again, because the estimated time may change at any time, For example, the initial estimated duration is 3 minutes, but during the 3 minutes, the customer service may leave for something, which will lengthen the waiting time, or if more customer service staff are added, the estimated time will become shorter. Recalculating based on the triggering of several events and obtaining more real-time features can make the estimated duration more accurate.
  • event operations such as click, slide, return, switch screen, etc.
  • the queue time estimation model after using the queue time estimation model to predict the queue time of the user, it may also include: in response to detecting that a customer service ends the session being processed and releases resources, taking out the queued user from the waiting queue and assigning it to the customer service, refreshing the waiting queue and updating the customer service status; when the customer service status is still that all customer services are busy, re-acquiring the queue characteristics of the user and the customer service status characteristics of the skill group, the customer service status characteristics include the current number of available customer services and the number of sessions being processed by each customer service; according to the queue characteristics of the user, the current number of available customer services and the number of sessions being processed by each customer service, using the queue time estimation model to predict the queue time of the user, so as to update the queue time of the user.
  • a customer service When a customer service ends a session and releases resources, it will trigger the system to update the waiting queue information and customer service status information, and assign a new online customer service consultation task to the customer service.
  • the system By updating the customer service status information and the waiting queue information, more accurate features can be provided to the algorithm, making the predicted queue time more accurate.
  • the method may further include: when the customer service status is that there is a customer service that is not busy, assigning the user to the customer service that is not busy.
  • the customer service status is that there is a customer service that is not busy
  • assigning the user to the customer service that is not busy When there is a customer service that is not busy, the user's online consultation request can be directly processed.
  • the user when the user is in the queue, he can also actively cancel the queue, at which point the process ends and there is no need to calculate the user's queue time.
  • Figure 2 is a schematic diagram of the prediction process of the online customer service consultation queue length in an embodiment of the present disclosure.
  • the prediction process of the online customer service consultation queue length is as follows: when receiving the user's online customer service consultation request, it is determined that the customer is online, and then the customer service status of the skill group is obtained, and whether there is an idle customer service is determined according to the customer service status; if there is an idle customer service, directly enter the online customer service consultation; otherwise, wait in the queue. Then, obtain the user's queue characteristics and the customer service status characteristics of the skill group, and use the queue length
  • the estimation model predicts the user's waiting time and outputs the predicted waiting time.
  • the user After that, it will monitor whether the user has triggered the specified operation event. If the user has triggered the specified operation event, it will re-execute the steps of obtaining the user's queue characteristics and the customer service status characteristics of the skill group, and using the queue time estimation model to predict the user's queue time, and output the predicted queue time; otherwise, it will continue to monitor whether there are customer service staff releasing resources. If a customer service staff releases resources, the queued user is taken out of the waiting queue and assigned to the customer service staff, and then the waiting queue is refreshed and the customer service status is updated. It will also re-execute the steps of obtaining the user's queue characteristics and the customer service status characteristics of the skill group, and using the queue time estimation model to predict the user's queue time, and output the predicted queue time.
  • Figure 3 is a schematic diagram of the error of the prediction result of the queue time in the embodiment of the present disclosure. As shown in Figure 3, the horizontal axis shows the error result in seconds, and the vertical axis shows the percentile value.
  • the prediction algorithm of the queue time in the embodiment of the present disclosure is used for prediction, and the average error of the prediction result is only 18.3 seconds. Among them, the prediction result error of the 25% percentile value is 2.8 seconds; the prediction result error of the 50% percentile value is 6.5 seconds; the prediction result error of the 75% percentile value is 15 seconds; the prediction result error of the 90% percentile value is 35.6 seconds; the prediction result error of the 99% percentile value is 212 seconds.
  • FIG4 is a schematic diagram of the main modules of the device for predicting the online customer service consultation queue length according to an embodiment of the present disclosure.
  • the device for predicting the online customer service consultation queue length 400 of the embodiment of the present disclosure mainly includes:
  • the status acquisition module 401 is used to determine the skill group corresponding to the user in response to receiving the user's online customer service consultation request, and obtain the customer service status of the skill group;
  • a queue adding module 402 configured to add the user to a waiting queue when the customer service status is that all customer service staff are busy;
  • the feature acquisition module 403 is used to acquire the queuing features of the user and the customer service status features of the skill group, wherein the customer service status features include the number of currently available customer service staff and the number of active customer service staff of each customer service staff. Number of sessions being processed;
  • the duration prediction module 404 is used to predict the queuing time of the user using a queuing time estimation model according to the queuing characteristics of the user, the number of currently available customer service representatives, and the number of sessions being processed by each customer service representative.
  • the queuing feature of the user includes the queue number of the user in the waiting queue and the queue number of the user in the skill group.
  • the duration prediction module 404 can also be used to: predict the queuing duration of the user using a queuing duration prediction model according to the queue number of the user in the waiting queue and the queue number of the user in the skill group, the number of currently available customer service representatives, and the number of sessions being processed by each customer service representative.
  • the device 400 for predicting the queuing time of online customer service consultation also includes a queuing time period determination module (not shown in the figure), which is used to: determine the queuing time period corresponding to the user according to the sending time of the user's online customer service consultation request before using the queuing time estimation model to predict the queuing time of the user according to the queuing characteristics of the user, the currently available number of customer services and the number of sessions being processed by each customer service; the time prediction module 404 can also be used to: use the queuing time estimation model to predict the queuing time of the user according to the queuing time period corresponding to the user, the queue number of the user in the waiting queue and the queue number of the user in the skill group, the currently available number of customer services and the number of sessions being processed by each customer service.
  • a queuing time period determination module (not shown in the figure), which is used to: determine the queuing time period corresponding to the user according to the sending time of the user's online customer service consultation request before using the que
  • the queuing time estimation model is constructed in the following manner: obtaining user online customer service consultation session data in a recent period; extracting session features corresponding to each session from the user online customer service consultation session data, the session features including the time of requesting online consultation, queuing time period, skill group, the queue number of the user in the waiting queue and the queue number of the user in the skill group, the number of currently available customer services of the skill group and the number of sessions being processed by each customer service, as well as the actual online consultation time; obtaining the queuing time corresponding to each session according to the time of requesting online consultation and the actual time of online consultation corresponding to each session; for each session, using the queue time corresponding to the session The queue duration marks the session features corresponding to the session to generate training data corresponding to the session; and performs model training based on the training data corresponding to each session to construct the queue duration estimation model.
  • the device 400 for predicting the queuing time of online customer service consultation also includes a queuing time updating module (not shown in the figure), which is used to: after using the queuing time prediction model to predict the queuing time of the user, in response to the user's specified operation trigger event, re-acquire the user's queuing characteristics and the customer service status characteristics of the skill group, the customer service status characteristics including the current number of available customer services and the number of sessions being processed by each customer service; based on the user's queuing characteristics, the current number of available customer services and the number of sessions being processed by each customer service, use the queuing time prediction model to predict the user's queuing time to update the user's queuing time.
  • a queuing time updating module (not shown in the figure), which is used to: after using the queuing time prediction model to predict the queuing time of the user, in response to the user's specified operation trigger event, re-acquire the user's queuing characteristics and the customer service
  • the device 400 for predicting the queuing time of online customer service consultation also includes a queuing time updating module (not shown in the figure), which is used to: after using the queuing time prediction model to predict the queuing time of the user, in response to detecting that a customer service has ended the session being processed and released resources, take out the queued user from the waiting queue and assign it to the customer service, refresh the waiting queue and update the customer service status; when the customer service status is still that all customer services are busy, re-acquire the queuing characteristics of the user and the customer service status characteristics of the skill group, the customer service status characteristics including the current number of available customer services and the number of sessions being processed by each customer service; based on the queuing characteristics of the user, the current number of available customer services and the number of sessions being processed by each customer service, use the queuing time prediction model to predict the queuing time of the user to update the queuing time of the user.
  • a queuing time updating module (not shown in the figure), which is used to:
  • the device 400 for predicting the queuing time of online customer service consultation also includes a customer service allocation module (not shown in the figure), which is used to: when the customer service status is that there is a customer service who is not busy, assign the user to the customer service who is not busy.
  • a customer service allocation module (not shown in the figure), which is used to: when the customer service status is that there is a customer service who is not busy, assign the user to the customer service who is not busy.
  • the invention in response to receiving the user's online customer
  • the invention provides a technical solution for predicting the queuing time of a user by using a queuing time estimation model based on the queuing characteristics of the user, the current number of available customer services and the number of sessions being processed by each customer service, and the queuing characteristics of the user.
  • the technical solution can predict the queuing time of the current user based on the status of online customer services that can receive the user in the customer service system at the time when the current user comes online, the status of the sessions being received by these customer services, and the queuing characteristics of the user, thereby greatly reducing the error of the prediction result and improving the prediction accuracy and performance of the queuing time.
  • FIG. 5 shows an exemplary system architecture 500 to which the method for predicting the online customer service consultation queue length or the device for predicting the online customer service consultation queue length according to the embodiments of the present disclosure can be applied.
  • system architecture 500 may include terminal devices 501, 502, 503, a network 504 and a server 505.
  • Network 504 is used to provide a medium for communication links between terminal devices 501, 502, 503 and server 505.
  • Network 504 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.
  • terminal devices 501, 502, 503 Users can use terminal devices 501, 502, 503 to interact with server 505 through network 504 to receive or send messages, etc.
  • Various communication client applications can be installed on terminal devices 501, 502, 503, such as shopping applications, web browser applications, customer service consulting applications, instant messaging tools, social platform software, etc. (only as examples).
  • the terminal devices 501 , 502 , and 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.
  • Server 505 may be a server that provides various services, such as a backend management server that supports shopping websites browsed by users using terminal devices 501, 502, and 503 (for example only).
  • the backend management server may provide services such as online customer service consultation requests received.
  • the data responds to receiving an online customer service consultation request from a user, determines the skill group corresponding to the user, and obtains the customer service status of the skill group; when the customer service status is that all customer services are busy, adds the user to the waiting queue; obtains the queuing characteristics of the user and the customer service status characteristics of the skill group, the customer service status characteristics including the current number of available customer services and the number of sessions being processed by each customer service; uses a queuing time estimation model to predict the queuing time of the user based on the queuing characteristics of the user, the current number of available customer services, and the number of sessions being processed by each customer service, and feeds back the processing results (such as the user's queuing time - just an example) to the terminal device.
  • the method for predicting the online customer service consultation queue length provided in the embodiment of the present disclosure is generally executed by the server 505 , and accordingly, the device for predicting the online customer service consultation queue length is generally set in the server 505 .
  • terminal devices, networks and servers in Figure 5 is only illustrative. Any number of terminal devices, networks and servers may be provided according to implementation requirements.
  • FIG. 6 a schematic diagram of a computer system 600 suitable for implementing a terminal device or server of the present disclosure is shown.
  • the terminal device or server shown in Figure 6 is only an example and should not limit the functions and scope of use of the present disclosure.
  • the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603.
  • ROM read-only memory
  • RAM random access memory
  • various programs and data required for the operation of the system 600 are also stored.
  • the CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604.
  • An input/output (I/O) interface 605 is also connected to the bus 604.
  • the I/O interface 605 includes an input section 606 including a keyboard, a mouse, etc.; a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.
  • the I/O interface 605 includes an output section 607; a storage section 608 including a hard disk or the like; and a communication section 609 including a network interface card such as a LAN card, a modem, or the like.
  • the communication section 609 performs communication processing via a network such as the Internet.
  • a drive 610 is also connected to the I/O interface 605 as needed.
  • a removable medium 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like, is installed on the drive 610 as needed so that a computer program read therefrom is installed into the storage section 608 as needed.
  • an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart.
  • the computer program can be downloaded and installed from a network through a communication part 609, and/or installed from a removable medium 611.
  • CPU central processing unit
  • the computer-readable medium shown in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above.
  • Computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.
  • a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or device.
  • a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above.
  • a computer-readable signal medium may also be any medium other than a computer-readable storage medium.
  • Computer-readable media that can send, propagate or transmit programs for use by or in conjunction with an instruction execution system, apparatus or device.
  • the program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.
  • each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function.
  • the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved.
  • each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
  • the units or modules involved in the embodiments described in the present disclosure may be implemented by software or by hardware.
  • the units or modules described may also be arranged in a processor.
  • a processor including a state acquisition module, a queue addition module, a feature acquisition module, and a duration prediction module.
  • the names of these units or modules do not constitute a limitation on the units or modules themselves under certain circumstances.
  • the duration prediction module may also be described as "a module for predicting the queuing duration of the user using a queuing duration estimation model based on the queuing characteristics of the user, the number of currently available customer services, and the number of sessions being processed by each customer service.”
  • the present disclosure further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device.
  • the above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device The method comprises: in response to receiving an online customer service consultation request from a user, determining the skill group corresponding to the user and obtaining the customer service status of the skill group; when the customer service status is that all customer service staff are busy, adding the user to a waiting queue; obtaining the queuing characteristics of the user and the customer service status characteristics of the skill group, the customer service status characteristics including the current number of available customer service staff and the number of sessions being processed by each customer service; and predicting the queuing time of the user using a queuing time estimation model according to the queuing characteristics of the user, the current number of available customer service staff and the number of sessions being processed by each customer service.
  • the skill group corresponding to the user is determined, and the customer service status of the skill group is obtained; when the customer service status is that all customer services are busy, the user is added to the waiting queue; the user's queuing characteristics and the customer service status characteristics of the skill group are obtained, and the customer service status characteristics include the current number of available customer services and the number of sessions being processed by each customer service; according to the user's queuing characteristics, the current number of available customer services and the number of sessions being processed by each customer service, a queuing time estimation model is used to predict the user's queuing time.
  • the technical solution can predict the current user's queuing time based on the status of online customer services that can receive the user in the customer service system at the time when the current user comes online, the status of sessions being received by these customer services, and the user's queuing characteristics, thereby greatly reducing the error of the prediction result and improving the prediction accuracy and prediction performance of the queuing time.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Business, Economics & Management (AREA)
  • Human Computer Interaction (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Finance (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Telephonic Communication Services (AREA)

Abstract

本公开提供了一种在线客服咨询排队时长的预测方法和装置，涉及计算机技术领域。该方法的一具体实施方式包括：响应于接收到用户的在线客服咨询请求，确定所述用户对应的技能组，并获取所述技能组的客服状态；在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中；获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。该实施方式大大降低了预测结果误差，提高了排队时长的预测准确度和预测性能。

Description

在线客服咨询排队时长的预测方法和装置
相关申请的交叉引用
本公开要求享有2023年6月13日提交的申请号为202310699670.6的中国发明专利申请的优先权，其全部内容通过引用并入本文。
技术领域
本公开涉及计算机技术领域，尤其涉及一种在线客服咨询排队时长的预测方法和装置。
背景技术
随着互联网电子商务的蓬勃发展，在线客服系统已成为电子商务网站的重要组成部分，作为消费者，通过在线客服系统咨询了解想要购买的商品是最便捷的途径。然而在线客服人力资源有限，忙时用户流量爆线，客服接待压力较大，导致部分用户无法及时进入人工客服，从而进入等待队列等候。在用户进入等待队列后，其核心就是要告知用户预计等待时间，这直接决定了用户是否选择进行排队，而排队时长预估的准确性，也直接影响用户的排队体验。
目前，在预估用户的排队时长时，一般可统计当前排队人数，通过排队进线平均时间乘以人数计算预估排队时长；也可基于NLP(Natural Language Processing，自然语言处理)技术预测单通会话结束事件，累加可接待该用户的所有客服正在进行中的所有会话的预计结束时间，得到该用户需要的排队时长。
在实现本公开过程中，发明人发现现有技术无论是基于统计的方法预估排队时长，还是基于NLP技术来预估排队时长，对排队时长的预估误差均较大，且预测性能较低。
发明内容
有鉴于此，本公开实施例提供一种在线客服咨询排队时长的预测方法和装置，大大降低了预测结果误差，提高了排队时长的预测准确度和预测性能。
为实现上述目的，根据本公开实施例的一个方面，提供了一种在线客服咨询排队时长的预测方法，包括：
响应于接收到用户的在线客服咨询请求，确定所述用户对应的技能组，并获取所述技能组的客服状态；
在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中；
获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；
根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
可选地，所述用户的排队特征包括所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，包括：根据所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
可选地，在根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长之前，还包括：根据所述用户的在线客服咨询请求的发送时间确定所述用户对应的排队时段；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，包括：根据所述用户对应的排队时段、所述用户在所述等待队列中的排队号和所述用户在所述技能组中 的排队号、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
可选地，所述排队时长预估模型是通过以下方式构建的：获取最近时段内的用户在线客服咨询会话数据；从所述用户在线客服咨询会话数据中提取每个会话对应的会话特征，所述会话特征包括请求进线咨询时间、排队时段、所属技能组、用户在等待队列中的排队号和用户在所述所属技能组中的排队号、所述所属技能组的当前可用客服数和每个客服正在处理的会话数，以及实际进线咨询时间；根据每个会话对应的请求进线咨询时间和实际进线咨询时间得到每个会话对应的排队时长；对每个会话，使用所述会话对应的排队时长对所述会话对应的会话特征进行打标以生成所述会话对应的训练数据；基于每个会话对应的训练数据来进行模型训练以构建所述排队时长预估模型。
可选地，在使用排队时长预估模型预测所述用户的排队时长之后，还包括：响应于所述用户的指定操作触发事件，重新获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，以更新所述用户的排队时长。
可选地，在使用排队时长预估模型预测所述用户的排队时长之后，还包括：响应于检测到有客服结束正在处理的会话并释放资源，从所述等待队列中取出排队用户并分配给所述客服，刷新所述等待队列并更新客服状态；在所述客服状态仍为所有客服均忙碌的情况下，重新获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，以更新所述用户的排队时长。
可选地，所述方法还包括：在所述客服状态为有客服未处于忙碌状态的情况下，将所述用户分配给所述未处于忙碌状态的客服。
根据本公开实施例的另一方面，提供了一种在线客服咨询排队时长的预测装置，包括：
状态获取模块，用于响应于接收到用户的在线客服咨询请求，确定所述用户对应的技能组，并获取所述技能组的客服状态；
队列添加模块，用于在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中；
特征获取模块，用于获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；
时长预测模块，用于根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
根据本公开实施例的又一方面，提供了一种电子设备，包括：一个或多个处理器；存储装置，用于存储一个或多个程序，当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现本公开实施例所提供的在线客服咨询排队时长的预测方法。
根据本公开实施例的再一方面，提供了一种计算机可读介质，其上存储有计算机程序，所述程序被处理器执行时实现本公开实施例所提供的在线客服咨询排队时长的预测方法。
上述公开中的一个实施例具有如下优点或有益效果：通过响应于接收到用户的在线客服咨询请求，确定用户对应的技能组，并获取技能组的客服状态；在客服状态为所有客服均忙碌的情况下，将用户添加到等待队列中；获取用户的排队特征和技能组的客服状态特征，客 服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据用户的排队特征、当前可用客服数和每个客服正在处理的会话数，使用排队时长预估模型预测用户的排队时长的技术方案，可以根据当前用户进线时刻客服系统中可接待该用户的在线客服状态、这些客服正在接待的会话状态，以及该用户的排队特征，预测当前用户的排队时长，大大降低了预测结果误差，提高了排队时长的预测准确度和预测性能。
上述的非惯用的可选方式所具有的进一步效果将在下文中结合具体实施方式加以说明。
附图说明
附图用于更好地理解本公开，不构成对本公开的不当限定。其中：
图1是根据本公开实施例的在线客服咨询排队时长的预测方法的主要步骤示意图；
图2是本公开一个实施例的在线客服咨询排队时长的预测流程示意图；
图3是本公开实施例的排队时长的预测结果误差示意图；
图4是根据本公开实施例的在线客服咨询排队时长的预测装置的主要模块示意图；
图5是本公开实施例可以应用于其中的示例性系统架构图；
图6是适于用来实现本公开实施例的终端设备或服务器的计算机系统的结构示意图。
具体实施方式
以下结合附图对本公开的示范性实施例做出说明，其中包括本公开实施例的各种细节以助于理解，应当将它们认为仅仅是示范性的。因此，本领域普通技术人员应当认识到，可以对这里描述的实施例做出各种改变和修改，而不会背离本公开的范围和精神。同样，为了清楚和简明，以下的描述中省略了对公知功能和结构的描述。
需要说明的是，本公开的技术方案中，所涉及的用户个人信息的采集、收集、更新、分析、处理、使用、传输、存储等方面，均符合相关法律法规的规定，被用于合法的用途，且不违背公序良俗。对用户个人信息采取必要措施，防止对用户个人信息数据的非法访问，维护用户个人信息安全、网络安全和国家安全。
随着互联网电子商务的蓬勃发展，在线客服系统已成为电子商务网站的重要组成部分，作为消费者，通过在线客服系统咨询了解想要购买的商品是最便捷的途径。然而在线客服人力资源和有限，客服人员一天的排班需要兼顾忙时和闲时，在忙时，用户流量爆线，客服接待压力较大，导致部分用户无法及时进入人工客服，从而进入排队等候队列。咨询排队即通过将咨询人工用户量削峰填谷，充分利用客服资源，让用户通过短暂的等待便可以进入人工客服，而不是直接走留言无客服回复，实现最大限度的提升用户接起率，提升用户体验。
在用户进入等待队列后，其核心就是要告知用户预计等待时间，这直接决定了用户是否选择进行排队，而排队时长预估的准确性，也直接影响用户的排队体验。因此，排队时长预估通过算法实现，基于历史客服结束会话的频次，以及预测每通正在进行的会话还有多长时间结束，以及不同用户在不同技能组下排队，综合计算用户的排队时长。电商客服系统中，自营商家少则几十多则几万客服需要同时进行接待，比较庞大的客服体系，拥有着数万客服需要同时管理和接待，客服个体加挂的技能组错中交错，往往一个客服需要接待几十个技能组的流量，因此无法通过简单的规则来确定用户进线需要排队多长时间。同时在线客服需要同时接待多通会话，每一通会话释放后才能接待下一个客服，每个用户因人而异，无法确定用户与客服的对话还有多久结束，因此要预估准确用户排队时长，是一个很难的课题。
目前在预估用户的排队时长时，多采用以下两种方式：(1)基于 统计的方法，统计当前排队人数，通过排队进线平均时间乘以人数即为预估等待时间；(2)基于NLP技术预测单通会话结束时间的方法，预测单通会话结束时间，累加可接待该用户的所有客服正在进行中的所有会话预计结束时间，得到单用户需要排队的时间。
然而，这两种方法都存在一定的缺陷，基于统计方法存在的问题主要如下：(1)每人进线排队时间差异较大，标准差较高，平均后导致整体预估时间误差较大；(2)无时段差异，无当前客服人数统计，例如凌晨时段客服数量较少，导致单用户排队进线时间较长，预测误差较大。基于NLP技术预测单通会话结束时间的方法存在的问题：(1)基于NLP技术根据当前用户与客服的对话内容，预测当通会话结束的时间，每一通会话预测时间误差较大，经测试发现无法多达几十秒，最终预估排队时间需要将多通会话预测时间累加，误差会无限放大；(2)需要预测当前在线客服所有正在接待的会话预计结束时间，一个用户排队时间预估需要执行成千上万次预测，性能较低，可行性低。
为了解决现有技术中存在的上述技术问题，本公开提供了一种在线客服咨询排队时长的预测方法，其实现原理是：根据当前用户进线时刻，获取当前客服系统中可接待该用户的在线客服状态，以及这些客服正在接待的会话状态，以及该用户所处等待队列的位置，提取相关的特征，同时将历史用户在当前特征下排队进线的真实排队时长作为标签，基于机器学习回归方法进行模型学习，预测当前用户进入等待队列时刻预计排队时长，将排队时长划分为多个分片，返回给前端呈现给用户。
图1是根据本公开实施例的在线客服咨询排队时长的预测方法的主要步骤示意图。如图1所示，本公开实施例的在线客服咨询排队时长的预测方法主要包括如下的步骤S101至步骤S104。
步骤S101：响应于接收到用户的在线客服咨询请求，确定所述用 户对应的技能组，并获取所述技能组的客服状态。当用户进线请求在线客服咨询时，系统会首先根据业务范围，确定该在线客服咨询请求对应的技能组，每个技能组对应一定的业务的在线客服咨询任务。
在确定该用户对应的技能组后，将获取该技能组的客服状态。每个技能组可能会包括多个客服，每个客服也可能会对应多个技能组，且每个客服可以同时处理多个在线客服咨询会话。当该技能组包括的所有客服均在处理会话，并且每个客服当前处理会话数量为该客服可处理的会话数量限值时，该技能组的客服状态为所有客服均忙碌。否则，若有客服未处理会话，或者有客服当前处理会话数量未达到该客服可处理的会话数量限值，则认为该技能组的客服状态为有客服空闲。
步骤S102：在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中。当该技能组的所有客服均在处理会话，且每个客服当前处理会话数量为该客服可处理的会话数量限值时，所有客服均忙碌，此时，将该用户添加到等待队列中。
步骤S103：获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数。在将用户添加到等待队列中之后，将根据用户的排队情况和该技能组的客服的状态特征来预估用户的排队时长。其中，用户的排队特征主要包括用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号。其中，该等待队列中保存了所有需要进行在线客服咨询的排队用户及其对应的在线客服咨询请求，并且，还保存了每个用户对应的技能组信息，故而，根据等待队列的信息即可获取用户在等待队列中的第一排队号和用户在对应的技能组中的第二排队号。
步骤S104：根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
根据本公开的一个实施例，所述用户的排队特征包括所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号。并且，根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，具体可以包括：根据所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
根据本公开的另一个实施例，在根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长之前，还可以包括：根据所述用户的在线客服咨询请求的发送时间确定所述用户对应的排队时段。并且，根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，具体可以包括：根据所述用户对应的排队时段、所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
根据本公开的实施例，所述排队时长预估模型是通过以下方式构建的：获取最近时段内的用户在线客服咨询会话数据；从所述用户在线客服咨询会话数据中提取每个会话对应的会话特征，所述会话特征包括请求进线咨询时间、排队时段、所属技能组、用户在等待队列中的排队号和用户在所述所属技能组中的排队号、所述所属技能组的当前可用客服数和每个客服正在处理的会话数，以及实际进线咨询时间；根据每个会话对应的请求进线咨询时间和实际进线咨询时间得到每个会话对应的排队时长；对每个会话，使用所述会话对应的排队时长对所述会话对应的会话特征进行打标以生成所述会话对应的训练数据；基于每个会话对应的训练数据来进行模型训练以构建所述排队时长预 估模型。
具体地，本公开在进行排队时长预估模型构建时，首先获取最近时段内的用户在线客服咨询会话数据，例如可筛选近一个月内用户排队进线的会话数据，客服接待数据，排队进线真实时长等数据。其中，用户排队进线的会话数据用于提取模型训练特征和标签；客服接待数据用于计算用户进线时刻、当前可接待该用户的客服数以及这些客服正在进行中的会话数；排队进线真实时长，通过对用户排队进线时刻进行埋点，获取用户从开始排队到进线与人工客服对话之间的时长。
本公开实施例为了构建某一时刻用户排队状态和客服接待状态与用户真实排队时长的映射关系，因此需要分别构建用户画像特征和客服画像特征，同时这些特征必需是在用户排队过程中的任意时刻可获取，特征定义如下：
排队时段：以小时分片，用户在该小时内进线的时段编号，因为在不同的时段，用户排队进线的时间是不同的，例如在凌晨进线排队，由于可接待的客服数较小，通常情况下排队进线时间会超过10分钟，因此排队时段可影响用户真实排队进线时长；
用户在等待队列中的排队号：统计应用客服系统中，当前用户所处在等待队列中的序号，排队序号越靠前，排队时间越短；
用户在所属技能组中的排队号：由于用户进线是先通过系统导航到相应的技能组下，在技能组下进行排队，因此需要记录在该技能组下的排队序号，如果总体排队人数较多，但技能组下排队人数较少，说明当前进线业务分布不均匀，在排队人数较少的技能组下仍然可以在较短的排队时间内进线；
所属技能组的当前可用客服数：即在客服系统中所有加挂了该用户进线技能组的客服数汇总，客服人数越多，说明释放正在进行中的会话速度越快，用户排队时间越短；
客服正在处理的会话数：可接待当前排队用户的客服正在同时接待中的会话数量，会话数越多，说明即将释放会话的概率越大，用户 排队时间越短；
所属技能组标识：用户进线是先通过系统导航到相应的技能组下，由于业务差异性和客服人员加挂的人数不同，不同的技能组排队进线的时间也有所不同，因此该特征也是作为预测用户排队时长的重要特征。
本公开实施例为了建立某一时刻用户排队状态和客服接待状态与用户真实排队时长的映射关系，需要构建用户真实排队时长的标签，用于模型学习。一般情况下，可以在用户开始排队和结束排队时刻分别埋点，记录用户真实的排队时长，即可作为模型训练数据的标签。然而，在未进行埋点时，则无法获取用户真实排队时间，因此需要构建一套模拟用户真实排队进线时长的方法，以此来作为模型学习的标签，具体方法如下：以分钟为粒度，取每一分钟内产生留言的会话，假设这些留言用户在一分钟的开始时刻同时进线产生排队，再取这一时刻正在发生的会话，获取每一通会话真正结束时间距离当前时刻的时长，按照时长从小到大进行排序，依次遍历排队用户与正在进行中的会话，若正在进行中的会话所接待的客服可接待排队用户技能组，则这通会话的结束时长即为当前排队用户的等待时长，依次类推。理论上，在客服未抢单、客服未上线和挂起、接线模式未调整上限等情况下，那么当前正在进行中的会话结束后，排队用户才有可能进线被接起。
下面举例说明如何计算用户的真实排队时长，如下表1中所示，其示出了本公开一个实施例的排队用户队列信息。
表1
如下的表2中示出了当前进行中的会话队列，主要包括客服编号(包括了客服挂载的技能组信息)、会话真实结束时长(单位为秒)、匹配队列用户索引几个字段。
表2
结合上述的表1和表2可以看出，0号排队用户所在技能组，当前 在线的客服均未加挂该技能组，该用户无法被接起，除非有可接待该技能组的客服上线才能接起该用户，这种情况应当提示用户排队时间较长或无可接待客服；1、2、3号用户均为此类情况。4号排队用户，在线客服中A4-1006客服可接待该技能组，因此该客服最近一通会话结束后，该排队用户即可进线，因此会话真实结束时长6秒即为该用户的排队等待时长。7号排队用户，在线客服中A3-1005客服可接待该技能组，因此该客服最近一通会话结束后，该排队用户即可进线，因此会话真实结束时长5秒即为该用户的排队时长。
通过以上的介绍，即可得到每个会话对应的排队时长。对每个会话，使用该会话对应的排队时长对该会话对应的会话特征进行打标以生成该会话对应的训练数据。之后，基于每个会话对应的训练数据来进行模型训练以构建排队时长预估模型。具体地，本公开的实施例中，在构建用户排队状态和客服接待状态与真实或模拟排队时长的映射关系时，可采用GBDT(Gradient Boosting Decision Tree)回归算法，基于前述的训练数据来进行排队时长预估模型的训练。
在构建排队时长预估模型之后，即可根据用户对应的排队时段、用户的排队特征、当前可用客服数和每个客服正在处理的会话数等特征来预测用户的排队时长。
根据本公开的又一个实施例，在使用排队时长预估模型预测所述用户的排队时长之后，还可以包括：响应于所述用户的指定操作触发事件，重新获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，以更新所述用户的排队时长。其中，指定操作例如是点击、滑动、返回、切屏等操作。当用户产生点击、滑动、返回、切屏等事件操作时，将会再次触发预测用户的排队时长，因为预估时间随时可能发生变化， 比如最开始预估时长为3分钟，但3分钟过程中有可能客服有事离开了，排队时间就会变长，或者增加了客服，预估时间就会变短，根据几种事件的触发来重新计算，获取更实时的特征，可以使得预估时长更准确。
根据本公开的又一个实施例，在使用排队时长预估模型预测所述用户的排队时长之后，还可以包括：响应于检测到有客服结束正在处理的会话并释放资源，从所述等待队列中取出排队用户并分配给所述客服，刷新所述等待队列并更新客服状态；在所述客服状态仍为所有客服均忙碌的情况下，重新获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，以更新所述用户的排队时长。当有客服结束了一个会话释放资源后，会触发系统更新等待队列信息和客服状态信息，并为该客服分配新的在线客服咨询任务。通过更新客服状态信息及等待队列信息，可以提供给算法更准确的特征，使得预测的排队时长更准确。
根据本公开的又一个实施例，所述方法还可以包括：在所述客服状态为有客服未处于忙碌状态的情况下，将所述用户分配给所述未处于忙碌状态的客服。当有客服不是处于忙碌状态时，则可直接对用户的在线咨询请求进行处理。另外，当用户在排队过程中，还可以主动取消排队，此时流程结束，无需再计算用户的排队时长。
图2是本公开一个实施例的在线客服咨询排队时长的预测流程示意图。如图2所示，本公开的实施例中，在线客服咨询排队时长的预测流程如下：当接收到用户的在线客服咨询请求，即判定客户进线，之后，获取技能组的客服状态，并根据客服状态判断是否有空闲客服；若有空闲客服，则直接进入在线客服咨询；否则，进行等待队列。然后，获取用户的排队特征和技能组的客服状态特征，并使用排队时长 预估模型预测用户的排队时长，输出预测的排队时长。
之后，将监测用户是否触发了指定操作事件，若用户触发了指定操作事件，则重新执行获取用户的排队特征和技能组的客服状态特征，并使用排队时长预估模型预测用户的排队时长，输出预测的排队时长的步骤；否则，持续监测是否有客服释放资源。若有客服释放资源，则从等待队列中取出排队用户并分配给该客服，然后刷新等待队列并更新客服状态，并再次执行获取用户的排队特征和技能组的客服状态特征，并使用排队时长预估模型预测用户的排队时长，输出预测的排队时长的步骤。
图3是本公开实施例的排队时长的预测结果误差示意图。如图3所示，横坐标显示为误差结果，单位为秒，纵坐标显示为分位值。使用本公开实施例的排队时长的预测算法进行预测，其预测结果误差均值仅为18.3秒。其中，25％分位值的预测结果误差为2.8秒；50％分位值的预测结果误差为6.5秒；75％分位值的预测结果误差为15秒；90％分位值的预测结果误差为35.6秒；99％分位值的预测结果误差212秒。
由此可以看出，使用本公开实施例的排队时长的预测算法进行排队时长预测，大大降低了预测结果误差，预测的排队时长更准确。
图4是根据本公开实施例的在线客服咨询排队时长的预测装置的主要模块示意图。如图4所示，本公开实施例的在线客服咨询排队时长的预测装置400主要包括：
状态获取模块401，用于响应于接收到用户的在线客服咨询请求，确定所述用户对应的技能组，并获取所述技能组的客服状态；
队列添加模块402，用于在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中；
特征获取模块403，用于获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正 在处理的会话数；
时长预测模块404，用于根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
根据本公开的一个实施例，所述用户的排队特征包括所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号。时长预测模块404还可以用于：根据所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
根据本公开的另一个实施例，在线客服咨询排队时长的预测装置400还包括排队时段确定模块(图中未示出)，用于：在根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长之前，根据所述用户的在线客服咨询请求的发送时间确定所述用户对应的排队时段；时长预测模块404还可以用于：根据所述用户对应的排队时段、所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
根据本公开的又一个实施例，所述排队时长预估模型是通过以下方式构建的：获取最近时段内的用户在线客服咨询会话数据；从所述用户在线客服咨询会话数据中提取每个会话对应的会话特征，所述会话特征包括请求进线咨询时间、排队时段、所属技能组、用户在等待队列中的排队号和用户在所述所属技能组中的排队号、所述所属技能组的当前可用客服数和每个客服正在处理的会话数，以及实际进线咨询时间；根据每个会话对应的请求进线咨询时间和实际进线咨询时间得到每个会话对应的排队时长；对每个会话，使用所述会话对应的排 队时长对所述会话对应的会话特征进行打标以生成所述会话对应的训练数据；基于每个会话对应的训练数据来进行模型训练以构建所述排队时长预估模型。
根据本公开的又一个实施例，在线客服咨询排队时长的预测装置400还包括排队时长更新模块(图中未示出)，用于：在使用排队时长预估模型预测所述用户的排队时长之后，响应于所述用户的指定操作触发事件，重新获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，以更新所述用户的排队时长。
根据本公开的又一个实施例，在线客服咨询排队时长的预测装置400还包括排队时长更新模块(图中未示出)，用于：在使用排队时长预估模型预测所述用户的排队时长之后，响应于检测到有客服结束正在处理的会话并释放资源，从所述等待队列中取出排队用户并分配给所述客服，刷新所述等待队列并更新客服状态；在所述客服状态仍为所有客服均忙碌的情况下，重新获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，以更新所述用户的排队时长。
根据本公开的又一个实施例，在线客服咨询排队时长的预测装置400还包括客服分配模块(图中未示出)，用于：在所述客服状态为有客服未处于忙碌状态的情况下，将所述用户分配给所述未处于忙碌状态的客服。
根据本公开实施例的技术方案，通过响应于接收到用户的在线客 服咨询请求，确定用户对应的技能组，并获取技能组的客服状态；在客服状态为所有客服均忙碌的情况下，将用户添加到等待队列中；获取用户的排队特征和技能组的客服状态特征，客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据用户的排队特征、当前可用客服数和每个客服正在处理的会话数，使用排队时长预估模型预测用户的排队时长的技术方案，可以根据当前用户进线时刻客服系统中可接待该用户的在线客服状态、这些客服正在接待的会话状态，以及该用户的排队特征，预测当前用户的排队时长，大大降低了预测结果误差，提高了排队时长的预测准确度和预测性能。
图5示出了可以应用本公开实施例的在线客服咨询排队时长的预测方法或在线客服咨询排队时长的预测装置的示例性系统架构500。
如图5所示，系统架构500可以包括终端设备501、502、503，网络504和服务器505。网络504用以在终端设备501、502、503和服务器505之间提供通信链路的介质。网络504可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。
用户可以使用终端设备501、502、503通过网络504与服务器505交互，以接收或发送消息等。终端设备501、502、503上可以安装有各种通讯客户端应用，例如购物类应用、网页浏览器应用、客服咨询类应用、即时通信工具、社交平台软件等(仅为示例)。
终端设备501、502、503可以是具有显示屏并且支持网页浏览的各种电子设备，包括但不限于智能手机、平板电脑、膝上型便携计算机和台式计算机等等。
服务器505可以是提供各种服务的服务器，例如对用户利用终端设备501、502、503所浏览的购物类网站提供支持的后台管理服务器(仅为示例)。后台管理服务器可以对接收到的在线客服咨询请求等 数据进行响应于接收到用户的在线客服咨询请求，确定所述用户对应的技能组，并获取所述技能组的客服状态；在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中；获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长等处理，并将处理结果(例如用户的排队时长--仅为示例)反馈给终端设备。
需要说明的是，本公开实施例所提供的在线客服咨询排队时长的预测方法一般由服务器505执行，相应地，在线客服咨询排队时长的预测装置一般设置于服务器505中。
应该理解，图5中的终端设备、网络和服务器的数目仅仅是示意性的。根据实现需要，可以具有任意数目的终端设备、网络和服务器。
下面参考图6，其示出了适于用来实现本公开实施例的终端设备或服务器的计算机系统600的结构示意图。图6示出的终端设备或服务器仅仅是一个示例，不应对本公开实施例的功能和使用范围带来任何限制。
如图6所示，计算机系统600包括中央处理单元(CPU)601，其可以根据存储在只读存储器(ROM)602中的程序或者从存储部分608加载到随机访问存储器(RAM)603中的程序而执行各种适当的动作和处理。在RAM 603中，还存储有系统600操作所需的各种程序和数据。CPU 601、ROM 602以及RAM 603通过总线604彼此相连。输入/输出(I/O)接口605也连接至总线604。
以下部件连接至I/O接口605：包括键盘、鼠标等的输入部分606；包括诸如阴极射线管(CRT)、液晶显示器(LCD)等以及扬声器等的 输出部分607；包括硬盘等的存储部分608；以及包括诸如LAN卡、调制解调器等的网络接口卡的通信部分609。通信部分609经由诸如因特网的网络执行通信处理。驱动器610也根据需要连接至I/O接口605。可拆卸介质611，诸如磁盘、光盘、磁光盘、半导体存储器等等，根据需要安装在驱动器610上，以便于从其上读出的计算机程序根据需要被安装入存储部分608。
特别地，根据本公开的实施例，上文参考流程图描述的过程可以被实现为计算机软件程序。例如，本公开的实施例包括一种计算机程序产品，其包括承载在计算机可读介质上的计算机程序，该计算机程序包含用于执行流程图所示的方法的程序代码。在这样的实施例中，该计算机程序可以通过通信部分609从网络上被下载和安装，和/或从可拆卸介质611被安装。在该计算机程序被中央处理单元(CPU)601执行时，执行本公开的系统中限定的上述功能。
需要说明的是，本公开所示的计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质或者是上述两者的任意组合。计算机可读存储介质例如可以是——但不限于——电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。计算机可读存储介质的更具体的例子可以包括但不限于：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本公开中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。而在本公开中，计算机可读的信号介质可以包括在基带中或者作为载波一部分传播的数据信号，其中承载了计算机可读的程序代码。这种传播的数据信号可以采用多种形式，包括但不限于电磁信号、光信号或上述的任意合适的组合。计算机可读的信号介质还可以是计算机可读存储介质以外的任何 计算机可读介质，该计算机可读介质可以发送、传播或者传输用于由指令执行系统、装置或者器件使用或者与其结合使用的程序。计算机可读介质上包含的程序代码可以用任何适当的介质传输，包括但不限于：无线、电线、光缆、RF等等，或者上述的任意合适的组合。
附图中的流程图和框图，图示了按照本公开各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分，上述模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图或流程图中的每个方框、以及框图或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
描述于本公开实施例中所涉及到的单元或模块可以通过软件的方式实现，也可以通过硬件的方式来实现。所描述的单元或模块也可以设置在处理器中，例如，可以描述为：一种处理器包括状态获取模块、队列添加模块、特征获取模块和时长预测模块。其中，这些单元或模块的名称在某种情况下并不构成对该单元或模块本身的限定，例如，时长预测模块还可以被描述为“用于根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长的模块”。
作为另一方面，本公开还提供了一种计算机可读介质，该计算机可读介质可以是上述实施例中描述的设备中所包含的；也可以是单独存在，而未装配入该设备中。上述计算机可读介质承载有一个或者多个程序，当上述一个或者多个程序被一个该设备执行时，使得该设备 包括：响应于接收到用户的在线客服咨询请求，确定所述用户对应的技能组，并获取所述技能组的客服状态；在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中；获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
根据本公开实施例的技术方案，通过响应于接收到用户的在线客服咨询请求，确定用户对应的技能组，并获取技能组的客服状态；在客服状态为所有客服均忙碌的情况下，将用户添加到等待队列中；获取用户的排队特征和技能组的客服状态特征，客服状态特征包括当前可用客服数和每个客服正在处理的会话数；根据用户的排队特征、当前可用客服数和每个客服正在处理的会话数，使用排队时长预估模型预测用户的排队时长的技术方案，可以根据当前用户进线时刻客服系统中可接待该用户的在线客服状态、这些客服正在接待的会话状态，以及该用户的排队特征，预测当前用户的排队时长，大大降低了预测结果误差，提高了排队时长的预测准确度和预测性能。
上述具体实施方式，并不构成对本公开保护范围的限制。本领域技术人员应该明白的是，取决于设计要求和其他因素，可以发生各种各样的修改、组合、子组合和替代。任何在本公开的精神和原则之内所作的修改、等同替换和改进等，均应包含在本公开保护范围之内。

Claims (10)

  1. 一种在线客服咨询排队时长的预测方法，其包括：
    响应于接收到用户的在线客服咨询请求，确定所述用户对应的技能组，并获取所述技能组的客服状态；
    在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中；
    获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；
    根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
  2. 根据权利要求1所述的方法，其中，所述用户的排队特征包括所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号；
    根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，包括：
    根据所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
  3. 根据权利要求2所述的方法，其中，在根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长之前，还包括：
    根据所述用户的在线客服咨询请求的发送时间确定所述用户对应的排队时段；
    根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，包括：
    根据所述用户对应的排队时段、所述用户在所述等待队列中的排队号和所述用户在所述技能组中的排队号、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
  4. 根据权利要求1-3中任一所述的方法，其中，所述排队时长预估模型是通过以下方式构建的：
    获取最近时段内的用户在线客服咨询会话数据；
    从所述用户在线客服咨询会话数据中提取每个会话对应的会话特征，所述会话特征包括请求进线咨询时间、排队时段、所属技能组、用户在等待队列中的排队号和用户在所述所属技能组中的排队号、所述所属技能组的当前可用客服数和每个客服正在处理的会话数，以及实际进线咨询时间；
    根据每个会话对应的请求进线咨询时间和实际进线咨询时间得到每个会话对应的排队时长；
    对每个会话，使用所述会话对应的排队时长对所述会话对应的会话特征进行打标以生成所述会话对应的训练数据；
    基于每个会话对应的训练数据来进行模型训练以构建所述排队时长预估模型。
  5. 根据权利要求1所述的方法，其中，在使用排队时长预估模型预测所述用户的排队时长之后，还包括：
    响应于所述用户的指定操作触发事件，重新获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；
    根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，以更新所述用户的排队时长。
  6. 根据权利要求1所述的方法，其中，在使用排队时长预估模型 预测所述用户的排队时长之后，还包括：
    响应于检测到有客服结束正在处理的会话并释放资源，从所述等待队列中取出排队用户并分配给所述客服，刷新所述等待队列并更新客服状态；
    在所述客服状态仍为所有客服均忙碌的情况下，重新获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；
    根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长，以更新所述用户的排队时长。
  7. 根据权利要求1所述的方法，其中，所述方法还包括：
    在所述客服状态为有客服未处于忙碌状态的情况下，将所述用户分配给所述未处于忙碌状态的客服。
  8. 一种在线客服咨询排队时长的预测装置，其包括：
    状态获取模块，用于响应于接收到用户的在线客服咨询请求，确定所述用户对应的技能组，并获取所述技能组的客服状态；
    队列添加模块，用于在所述客服状态为所有客服均忙碌的情况下，将所述用户添加到等待队列中；
    特征获取模块，用于获取所述用户的排队特征和所述技能组的客服状态特征，所述客服状态特征包括当前可用客服数和每个客服正在处理的会话数；
    时长预测模块，用于根据所述用户的排队特征、所述当前可用客服数和所述每个客服正在处理的会话数，使用排队时长预估模型预测所述用户的排队时长。
  9. 一种电子设备，其包括：
    一个或多个处理器；
    存储装置，用于存储一个或多个程序，
    当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如权利要求1-7中任一所述的方法。
  10. 一种计算机可读介质，其上存储有计算机程序，所述程序被处理器执行时实现如权利要求1-7中任一所述的方法。
PCT/CN2024/071118 2023-06-13 2024-01-08 在线客服咨询排队时长的预测方法和装置 Ceased WO2024255225A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202310699670.6A CN117131167A (zh) 2023-06-13 2023-06-13 在线客服咨询排队时长的预测方法和装置
CN202310699670.6 2023-06-13

Publications (1)

Publication Number Publication Date
WO2024255225A1 true WO2024255225A1 (zh) 2024-12-19

Family

ID=88853418

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/071118 Ceased WO2024255225A1 (zh) 2023-06-13 2024-01-08 在线客服咨询排队时长的预测方法和装置

Country Status (2)

Country Link
CN (1) CN117131167A (zh)
WO (1) WO2024255225A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117131167A (zh) * 2023-06-13 2023-11-28 北京沃东天骏信息技术有限公司 在线客服咨询排队时长的预测方法和装置
CN118486438B (zh) * 2024-05-21 2025-04-18 苍兰棱星(南京)信息技术有限公司 一种用于远程医疗服务的服务优化管理系统
CN118612344A (zh) * 2024-06-21 2024-09-06 点控云(北京)智能科技有限公司 一种在线客服用户进线排队与分配装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107483357A (zh) * 2017-08-11 2017-12-15 广东电网有限责任公司信息中心 一种客服负载均衡系统及方法
CN111988478A (zh) * 2020-09-02 2020-11-24 深圳壹账通智能科技有限公司 一种呼入管理方法、装置、服务器及存储介质
CN112633567A (zh) * 2020-12-16 2021-04-09 深圳前海微众银行股份有限公司 一种等待时长的预测方法、设备及存储介质
CN112686528A (zh) * 2020-12-28 2021-04-20 京东数字科技控股股份有限公司 用于分配客服资源的方法、装置、服务器和介质
CN114418396A (zh) * 2022-01-19 2022-04-29 平安壹钱包电子商务有限公司 客服资源路由方法、装置、设备及存储介质
CN117131167A (zh) * 2023-06-13 2023-11-28 北京沃东天骏信息技术有限公司 在线客服咨询排队时长的预测方法和装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110147905B (zh) * 2019-05-08 2022-03-25 联想(北京)有限公司 信息处理方法、装置、系统及存储介质
CN113240200A (zh) * 2021-06-07 2021-08-10 中国银行股份有限公司 一种线上客服调度方法及装置
CN114037146A (zh) * 2021-11-05 2022-02-11 北京市商汤科技开发有限公司 一种排队等待时长确定方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107483357A (zh) * 2017-08-11 2017-12-15 广东电网有限责任公司信息中心 一种客服负载均衡系统及方法
CN111988478A (zh) * 2020-09-02 2020-11-24 深圳壹账通智能科技有限公司 一种呼入管理方法、装置、服务器及存储介质
CN112633567A (zh) * 2020-12-16 2021-04-09 深圳前海微众银行股份有限公司 一种等待时长的预测方法、设备及存储介质
CN112686528A (zh) * 2020-12-28 2021-04-20 京东数字科技控股股份有限公司 用于分配客服资源的方法、装置、服务器和介质
CN114418396A (zh) * 2022-01-19 2022-04-29 平安壹钱包电子商务有限公司 客服资源路由方法、装置、设备及存储介质
CN117131167A (zh) * 2023-06-13 2023-11-28 北京沃东天骏信息技术有限公司 在线客服咨询排队时长的预测方法和装置

Also Published As

Publication number Publication date
CN117131167A (zh) 2023-11-28

Similar Documents

Publication Publication Date Title
WO2024255225A1 (zh) 在线客服咨询排队时长的预测方法和装置
CN109684358B (zh) 数据查询的方法和装置
CN105281981B (zh) 网络服务的数据流量监控方法和装置
CN111786895A (zh) 动态全局限流的方法和装置
US12177310B2 (en) Method and apparatus for processing notification trigger message
CN108897854A (zh) 一种超时任务的监控方法和装置
CN111131356B (zh) 用于生成信息的方法和装置
CN113760982A (zh) 一种数据处理方法和装置
US11750711B1 (en) Systems and methods for adaptively rate limiting client service requests at a blockchain service provider platform
CN108810047B (zh) 用于确定信息推送准确率的方法、装置及服务器
CN107612844A (zh) 一种减轻服务器脉冲压力的方法、服务器和客户端
CN118071361A (zh) 延时分配客服的方法、装置、电子设备和计算机可读介质
CN109428926A (zh) 一种调度任务节点的方法和装置
CN111339495B (zh) 直播间在线人数的统计方法、装置、电子设备及存储介质
US20250370814A1 (en) Resource scheduling method based on cloud storage system, electronic device, and storage medium
CN112783924B (zh) 一种脏数据识别方法、装置和系统
US20190109945A1 (en) Technologies for managing unresolved customer interactions
CN113238919A (zh) 一种用户访问数的统计方法、装置及系统
WO2025035872A1 (zh) 数据库管理方法和装置
CN111786801A (zh) 一种基于数据流量进行计费的方法和装置
CN115396434B (zh) 一种消息处理的方法和装置
CN108564406A (zh) 一种激励推送的方法和装置
CN115801763A (zh) 文件传输方法、装置、电子设备及存储介质
CN115456229A (zh) 一种网点预约方法及装置
CN115729686A (zh) 一种数据处理方法、装置、设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 24822186

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE