CN111260102A

CN111260102A - User satisfaction prediction method and device, electronic equipment and storage medium

Info

Publication number: CN111260102A
Application number: CN201811459764.1A
Authority: CN
Inventors: 张坤雷
Original assignee: Beijing Didi Infinity Technology and Development Co Ltd
Current assignee: Beijing Didi Infinity Technology and Development Co Ltd
Priority date: 2018-11-30
Filing date: 2018-11-30
Publication date: 2020-06-09

Abstract

The application relates to the technical field of artificial intelligence, in particular to a user satisfaction prediction method, which comprises the following steps: acquiring user attribute information of a target user, current service information provided by a platform for the target user and customer service attribute information of customer service serving the target user; performing characteristic value processing on the acquired user attribute information of the target user, current service information provided by the platform for the target user and customer service attribute information of customer service serving the target user, and determining a model input characteristic corresponding to the target user; and inputting the determined model input characteristics into a pre-trained user satisfaction prediction model to obtain a user satisfaction prediction result of the target user. By adopting the method, the satisfaction degree of the target user can be predicted, and the predicted coverage rate and efficiency are high. The application also provides a user satisfaction prediction device, electronic equipment and a storage medium.

Description

User satisfaction prediction method and device, electronic equipment and storage medium

Technical Field

The present application relates to the field of artificial intelligence technologies, and in particular, to a method and an apparatus for predicting user satisfaction, an electronic device, and a storage medium.

Background

With the increasing requirements of users on service quality, how to improve the satisfaction of users is more and more emphasized. The net car appointment travel field is used as a new service application field in the present generation, and a net car appointment travel platform is used for receiving a car-taking request issued by a passenger through a passenger terminal and pushing an order generated according to the car-taking request to a driver terminal, so that the driver can receive an order at the driver terminal and directly communicate with the passenger to realize the net car appointment travel, historical services are often analyzed by analyzing the satisfaction degree of a net car appointment user (namely the passenger or the driver), and further, the service quality is optimized.

During the entire taxi taking process, situations may arise in which the passengers are dissatisfied, such as: the problems of driver detour, poor driver attitude, bad vehicle condition and the like may also occur, and the situation that the driver is not satisfied, such as: the passengers do not take the bus according to appointed time, the destination required by the passengers is not consistent with the destination of the order, and the like, and at the moment, the network appointment user can complain to the customer service. The customer service is used as an important interface for connecting a travel platform and a network car booking user, and is an important channel for acquiring user voice and processing actual problems, and the satisfaction degree of the user on the customer service is an important reference for measuring the platform service quality and optimizing the platform service quality.

In the related art, evaluation of satisfaction usually depends on investigation, for example, questionnaire links are published on the internet or investigation is performed in a manner of asking for scoring after a telephone operator service handles a problem, but the evaluation efficiency in the investigation manner is low, and the investigation recovery amount is small, which often has no statistical significance and results are low in referential property.

Disclosure of Invention

In view of this, an object of the embodiments of the present application is to provide a user satisfaction prediction method, an apparatus, an electronic device, and a storage medium, so as to predict the satisfaction of a target user, where both the coverage rate and the efficiency of prediction are high.

Mainly comprises the following aspects:

in a first aspect, an embodiment of the present application provides a user satisfaction prediction method, where the method includes:

acquiring user attribute information of a target user, current service information provided by a platform for the target user and customer service attribute information of customer service serving the target user;

performing characteristic value processing on the acquired user attribute information of the target user, current service information provided by the platform for the target user and customer service attribute information of customer service serving the target user, and determining a model input characteristic corresponding to the target user;

and inputting the determined model input characteristics into a pre-trained user satisfaction prediction model to obtain a user satisfaction prediction result of the target user.

In some embodiments, the current service information includes recorded content of a current customer service order; the determining of the model input features corresponding to the target user includes:

dividing the text content of the obtained record content of the current customer service work order;

inputting the text content into a pre-trained feature construction model aiming at each divided text content to obtain a feature vector corresponding to each text content;

and forming a characteristic vector group by using the characteristic vectors corresponding to all the divided text contents, and using the characteristic vector group as the model input characteristic corresponding to the target user.

In some embodiments, the current service information includes attribute information of a current customer service order; wherein, the attribute information of the current customer service work order comprises: whether the waiting time is the work order of repeated incoming lines and the interactive voice response IVR waiting time;

the determining of the model input features corresponding to the target user includes:

and using characteristic values respectively representing whether the work order is a repeated incoming line or not and the IVR waiting time as the model input characteristics.

In some embodiments, the current service information comprises current order information; wherein the current order information comprises: whether to automatically judge responsibility and the waiting time of the user;

and taking the characteristic values respectively representing whether the automatic accountability judgment is carried out and the waiting time of the user as the model input characteristics.

In one embodiment, the method further comprises:

and comparing the determined user satisfaction prediction result with a preset satisfaction threshold, judging whether the target user is satisfied with the customer service serving the target user, and if the target user is not satisfied with the customer service, pushing customer dissatisfaction prompt information to a responsible person.

In another embodiment, the current service information provided by the platform for the target user includes the recorded content of the customer service work order; when the target user is determined to be not satisfied with the customer service of the service, the method further comprises the following steps:

determining the latest historical order information of the target user corresponding to the customer service work order;

according to multiple preset dissatisfaction reason categories, extracting order information concerned by each dissatisfaction reason category from the determined historical order information;

and comparing the extracted order information concerned by each dissatisfaction reason category with dissatisfaction order information corresponding to the dissatisfaction reason category to determine whether the customer service order conforms to the dissatisfaction reason category.

In another embodiment, upon determining that the target user is not satisfied with the customer service serving the target user, the method further comprises:

and performing characteristic value processing on the acquired current service information provided by the platform for the target user, and inputting the processed characteristic value into a pre-trained dissatisfaction classification model to obtain a dissatisfaction type prediction result of the target user on customer service.

In some embodiments, the unsatisfactory classification model is trained according to the following steps:

acquiring the record content of the historical customer service work order corresponding to each sample user;

extracting dissatisfaction label classification information of the sample user from the acquired record content of the historical customer service work order;

and taking the recorded content of the historical customer service work order as an influence factor of unsatisfied classification, taking the classification information of the unsatisfied label as an unsatisfied classification result, and training to obtain the unsatisfied classification model.

In yet another embodiment, the user satisfaction prediction model is trained as follows:

acquiring user attribute information of each sample user, historical service information provided by a platform for the sample user, customer service attribute information of customer service serving the sample user, and satisfaction scoring information of the sample user for the customer service;

and taking the user attribute information, the historical service information and the customer service attribute information as user satisfaction influence factors, taking the satisfaction scoring information as a user satisfaction result, and training to obtain a user satisfaction prediction model.

In some embodiments, the training of the user satisfaction prediction model using the user attribute information, the historical service information, and the customer service attribute information as user satisfaction influencing factors and the satisfaction scoring information as user satisfaction results includes:

carrying out characteristic value processing on the user satisfaction influence factors, and determining the model input characteristics corresponding to the sample user;

and performing at least one round of model training by taking the model input characteristics as independent variables and the satisfaction degree scoring information as dependent variables to obtain the user satisfaction degree prediction model.

In a second aspect, an embodiment of the present application provides an apparatus for predicting user satisfaction, where the apparatus includes:

the information acquisition module is used for acquiring user attribute information of a target user, current service information provided by the platform for the target user and customer service attribute information of customer service serving the target user;

the characteristic determining module is used for performing characteristic value processing on the acquired user attribute information of the target user, the current service information provided by the platform for the target user and the customer service attribute information of the customer service serving the target user to determine the model input characteristic corresponding to the target user;

and the satisfaction prediction module is used for inputting the determined model input characteristics into a pre-trained user satisfaction prediction model to obtain a user satisfaction prediction result of the target user.

In some embodiments, the current service information includes recorded content of a current customer service order; the feature determination module is specifically configured to:

the feature determination module is specifically configured to:

In one embodiment, the apparatus further comprises:

and the judging module is used for comparing the determined user satisfaction prediction result with a preset satisfaction threshold value, judging whether the target user is satisfied with the customer service serving the target user, and if the target user is not satisfied with the customer service, pushing customer dissatisfaction prompt information to a responsible person.

In another embodiment, the current service information provided by the platform for the target user includes the recorded content of the customer service work order; the device further comprises:

the unsatisfied category determining module is used for determining the latest historical order information of the target user corresponding to the customer service work order;

In yet another embodiment, the apparatus further comprises:

and the unsatisfied type prediction module is used for carrying out characteristic value processing on the acquired current service information provided by the platform for the target user, and inputting the processed characteristic value into a pre-trained unsatisfied classification model to obtain an unsatisfied type prediction result of the target user on customer service.

In yet another embodiment, the apparatus further comprises:

the classification model training module is used for acquiring the record content of the historical customer service work order corresponding to each sample user;

In some embodiments, the apparatus further comprises:

the prediction model training module is used for acquiring user attribute information of each sample user, historical service information provided by the platform for the sample user, customer service attribute information of customer service serving the sample user and satisfaction scoring information of the sample user for the customer service;

In some embodiments, the predictive model training module is specifically configured to:

In a third aspect, an embodiment of the present application further provides an electronic device, including: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor and the storage medium communicating via the bus when the electronic device is running, the processor executing the machine-readable instructions to perform the steps of the user satisfaction prediction method according to the first aspect.

In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to perform the steps of the user satisfaction prediction method according to the first aspect.

By adopting the scheme, the obtained user attribute information, current service information and customer service attribute information related to the target user can be subjected to characteristic value processing, and then the processed model input characteristics are input into a trained user satisfaction prediction model so as to predict the satisfaction of the target user. That is, the satisfaction degree of the target user is predicted based on the user satisfaction degree prediction model obtained through training, the problems that the efficiency is low and the problem recovery amount is small due to the adoption of an investigation mode in the related technology are solved, the satisfaction degree of the target user can be predicted, the predicted coverage rate and the predicted efficiency are high, and further important reference is provided for the improvement of the platform service quality.

In addition, the method and the device for determining the pain point of the user can determine the category of the unsatisfied reasons when the target user is determined to be unsatisfied with the customer service of the service, can timely know the pain point of the user, facilitate follow-up of the pain point, solve the problem of the user and enable the user to be satisfied, and therefore improve the experience degree of the user.

In order to make the aforementioned objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.

Fig. 1 is a flowchart illustrating a user satisfaction prediction method according to an embodiment of the present application;

fig. 2 is a flowchart illustrating a user satisfaction predicting method according to a second embodiment of the present application;

fig. 3 is a flowchart illustrating a user satisfaction predicting method according to a fifth embodiment of the present application;

fig. 4 shows a flowchart of a user satisfaction prediction method provided in the sixth embodiment of the present application;

fig. 5 is a flowchart illustrating a user satisfaction predicting method according to a seventh embodiment of the present application;

fig. 6 is a schematic structural diagram illustrating a user satisfaction predicting apparatus according to an eighth embodiment of the present application;

fig. 7 shows a schematic structural diagram of an electronic device according to a ninth embodiment of the present application.

Detailed Description

In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it should be understood that the drawings in the present application are for illustrative and descriptive purposes only and are not used to limit the scope of protection of the present application. Additionally, it should be understood that the schematic drawings are not necessarily drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of the present application. It should be understood that the operations of the flow diagrams may be performed out of order, and steps without logical context may be performed in reverse order or simultaneously. One skilled in the art, under the guidance of this application, may add one or more other operations to, or remove one or more operations from, the flowchart.

In addition, the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application.

Considering that the evaluation of the satisfaction degree in the related art is usually performed by research, the research has low evaluation efficiency and the problem recovery amount is low, so that the pain point of the user cannot be known in time. In view of this, the present application provides a user satisfaction predicting method, an apparatus, an electronic device, and a storage medium, which may be applied to any scenario for predicting user satisfaction, where the user may be a passenger of a network appointment or a driver of the network appointment, and the network appointment may be a fast car, a special car, a tailgating car, a taxi, and the like, and the present application does not specifically limit this. This is described in more detail below by way of several examples.

Example one

As shown in fig. 1, a flowchart of a user satisfaction prediction method according to an embodiment of the present application is provided, where the user satisfaction prediction method includes the following steps:

s101, obtaining user attribute information of a target user, current service information provided by a platform for the target user, and customer service attribute information of customer service serving the target user.

Here, both the user attribute information of the target user and the customer service attribute information of the customer service serving the target user may belong to the inherent behavior characteristics of the person. In the embodiment of the application, the adopted user attribute information can be different when the satisfaction degree prediction is carried out on different users. For example, when the satisfaction degree of the user, i.e. the online car booking driver, is predicted, the user attribute information may include driver basic information including driver identification number (ID), driver telephone number, city ID of driver identity card, city ID of driver license plate, city ID of making order, maximum amount of robbing, city entrance, driver age, driver vehicle class, driver driving age, and the like, and may also include driver income information including income, subsidy rate, total reward, reward account income ratio, reward account ratio, reward day ratio, Gross transaction amount (Gross transaction Volume, GMV), Gross transaction amount (i.e. actual income per order), and may also include driver incoming line amount, Gross Interactive Voice Response (IVR) waiting time, Gross ringing time, Gross call time, repeat incoming line amount, pay incoming line amount, and pay incoming line amount, The incoming line information of the driver including the average incoming line time, the IVR hang-up times, the hang-up waiting times, the incoming line attenuation and the like. Similarly, when the satisfaction degree of the user, the car booking passenger, is predicted, the user attribute information may be driver basic information including information of the passenger ID, the passenger telephone number, the city ID to which the passenger identity card belongs, whether the city entrance is present, the passenger age, and the like, and may also be passenger incoming line information including passenger incoming line amount, list-average interactive voice response IVR waiting time, list-average ringing time, list-average call time, repeated incoming line amount, compensation incoming line amount, list-average incoming line time, IVR hang-up times, waiting hang-up times, incoming line attenuation, and the like. Corresponding to the user, the customer service attribute information of the customer service serving the user can be the basic customer service information including information of customer service ID, service line, customer service property, scene ID, native place, marital state, age, academic history, working age, employee state and the like, and can also be the customer service work information including information of bill-mean call duration, complaint establishment proportion, repeated line incoming quantity, compensation line incoming quantity, complaint quantity, bill-mean satisfaction degree, satisfaction degree parameter quantity attenuation, bill-mean satisfaction degree attenuation, compensation amount, satisfaction degree parameter quantity proportion and the like.

In addition, the current service information may only include current order information generated by the current order taking of the whole network car booking travel platform, may also only include relevant information of the current customer service work order generated when the user enters the line to the customer service according to whether the order is satisfied or not after the order is completed (such as attribute information describing the specific situation of the work order and record content when the customer service communicates with the user for specific problems), and may also include the current order information and relevant information of the current customer service work order. It should be noted that the current service information may only include attribute information and record content of the current customer service order, however, the current service information may not include current order information, and the current order information may not include the current order information as long as it is considered that an incoming line does not necessarily correspond to an order, for example, an application scenario in which a user consults a relevant service to the incoming line customer service.

The current order information may be order information including information of estimated distance, whether detour is performed, estimated price, waiting time, driving receiving distance, issuing and answering time difference, order cancellation type, responsibility judgment result, whether peak is performed in the morning and evening or not, the related information of the current customer service work order may be attribute information including information of work order state, processing person ID, processing department ID, ringing time, waiting time, IRV time, call time, work order detail problem description length, problem path name, work order problem description length and the like, and may be recording information including recording content of the customer service.

It should be noted that, the selection of each piece of information in the embodiment of the present application may have a certain pertinence, for example, the service working age may be used to describe the working life of the service, the side surface reflects the working experience of the service, the first order may describe the time when the driver/passenger uses the platform, the side surface reflects the experience of the driver/passenger on the platform, and the order information may be used to describe the situation of the driver's order, and may reflect the problem of incoming line and the possibility of satisfaction of the driver according to the situation of the order. For another example, for the current service information, whether the incoming line is a repeated incoming line or not can be laterally reflected whether the driver/passenger influences the incoming line because the incoming line is unsatisfactory last time or not, whether the IVR waiting time is too long or not influences the incoming line experience, and the like. It can be seen that the selection of the user attribute information, the customer service attribute information, the current service information, and the information included therein is important. In this embodiment of the application, the user attribute information may be part or all of the listed information, or may be user attribute information that includes both part or all of the listed information and other information, so as to improve the comprehensiveness of the data preparation stage, and further improve the accuracy and robustness of the subsequent training model. The description of the related customer service attribute information and the current service information is similar to that of the user attribute information, and is not repeated herein.

S102, performing characteristic value processing on the acquired user attribute information of the target user, current service information provided by the platform for the target user and customer service attribute information of customer service serving the target user, and determining a model input characteristic corresponding to the target user.

Here, after obtaining the user satisfaction influence factors (i.e., the user attribute information, the current service information, and the customer service attribute information) of the target user, the user satisfaction prediction method provided in the embodiment of the present application may perform feature value processing such as filtering, type conversion, derivation, and digitization on the user satisfaction influence factors, and may also perform feature value processing such as vector conversion on the user satisfaction influence factors, so as to obtain the processed user satisfaction influence factors.

The filtering processing refers to filtering operation on missing information, repeated information and the like in user satisfaction influence factors; the type conversion can be to normalize the influence factors of user satisfaction, so as to unify data from different sources under a reference system, and the comparison is meaningful, for example, for the IVR waiting time included in the attribute information of the current customer service order, the type conversion can normalize the IVR waiting time, for example, for the network appointment user waiting time included in the current order information, the type conversion can normalize the waiting time; the derivation processing refers to obtaining additional user satisfaction influence factors through statistical analysis according to the user satisfaction influence factors, for example, for the user satisfaction influence factors including driver basic information, relevant statistical information such as average ticket-grabbing times, ticket-grabbing income per time and the like of a certain driver can be obtained through derivation of the driver basic information; the above-mentioned digital processing may be to convert the user satisfaction influence factor into a number to represent, for example, for a work order of whether the incoming line is repeated, 1 (the incoming line is repeated) or 0 (the incoming line is not repeated) may be used to represent, for whether the responsibility is automatically determined, 1 (the responsibility is automatically determined) or 0 (the responsibility is not automatically determined), and so on.

In addition, the vector conversion processing may be vectorization of the user satisfaction influence factor, for example, for the record content of the current customer service work order, the text record may be encoded into a digital vector that can be recognized by the electronic device by means of natural language processing.

S103, inputting the determined model input characteristics into a pre-trained user satisfaction prediction model to obtain a user satisfaction prediction result of the target user.

Here, in the embodiment of the present application, a user satisfaction prediction result of a target user may be obtained by inputting, to a pre-trained user satisfaction prediction model, a model input feature corresponding to the target user, which is obtained by performing feature value processing on the acquired user attribute information of the target user, current service information provided by a platform for the target user, and customer service attribute information of customer service serving the target user. Therefore, the satisfaction degree prediction can be quickly and efficiently carried out on the target user by adopting the pre-trained user satisfaction degree prediction model, and the predicted coverage rate and the predicted efficiency are high. In the embodiment of the application, an integrated tree model, such as an Extreme Gradient Boosting tree (XGBOOST) model, may be used as the user satisfaction prediction model, and the model training stage is a process of training some unknown parameter information in the model.

It is worth noting that the satisfaction prediction result for the target user may be a predicted satisfaction likelihood size, which may be a percentage. In order to conveniently and visually display the satisfaction degree prediction result, the satisfaction degree score obtained by model prediction can be converted into the user satisfaction degree score based on the logistic regression function, so that the model is easy to display, the user satisfaction degree result corresponding to the model training stage can be backtracked, and the practicability is better.

In specific implementation, the user satisfaction influencing factors are various, and different feature processing methods can be adopted for different user satisfaction influencing factors. The obtained model input characteristics are different by adopting different characteristic processing methods. The determination of the corresponding model input features for different user satisfaction influencing factors is exemplified by the following embodiments two to four.

First, the process of determining the corresponding model input features by the current service information including the recorded content of the current customer service order is further described by the following embodiment two:

example two

As shown in fig. 2, a flowchart of a user satisfaction prediction method provided in the second embodiment is provided, where the method specifically includes the following steps:

s201, dividing the text content of the acquired record content of the current customer service work order;

s202, aiming at each divided text content, inputting the text content into a pre-trained feature construction model to obtain a feature vector corresponding to each text content;

and S203, forming a feature vector group by using the feature vectors corresponding to all the divided text contents, and using the feature vector group as the model input feature corresponding to the target user.

Here, it is considered that the recorded content of the current customer service order is taken as a special user satisfaction influence factor, which is in the form of text description, for which electronic equipment generally cannot be directly recognized. In order to solve the above problem, in the second embodiment of the present application, a text record is encoded into a digital vector that can be recognized by an electronic device by a natural language processing method to perform feature processing on the record content of the current customer service order.

After receiving the record content of the current customer service work order, text content division may be performed on the record content, and for each divided text content, the text content serving as a natural language may be converted into digital information in a vector form based on a word representation model, such as a word2vec model, so as to facilitate machine identification, which is called encoding. That is, a semantic feature vector is used to represent a text content (e.g., a word), and then feature vectors corresponding to all the divided text contents may be grouped into a feature vector group, and the feature vector group is used as a model input feature of the user satisfaction prediction model.

The above common word Representation models mainly include two types, One is a word Representation model based on One-hot coding Representation (One-hot coding Representation), and the other is a word Representation model based on distributed Representation (DistributedRepresentation).

The former word representation model uses a very long vector to represent a word, the length of the vector is the word size N of the dictionary, each vector only has one dimension of 1, the rest dimensions are all 0, and the position of 1 represents the position of the word in the dictionary. That is, the former word representation model stores word information in a sparse manner, that is, each word is assigned with a digital identifier, and the representation form is relatively simple. The latter word representation model needs to perform semantic representation according to context information, that is, words appearing in the same context have similar semantics, and thus the latter word representation model stores word information in a dense manner, and the representation form is relatively complex.

Considering that the former word Representation model based on One-hot Representation often encounters dimension disaster when solving practical problems and cannot reveal potential relations among vocabularies, the latter word Representation model based on Distributed Representation can be adopted to carry out vector Representation on tag information in specific implementation, so that the problem of dimension disaster is avoided, and correlation attributes among the vocabularies are mined, thereby improving the accuracy of semantic expression.

Next, the process of determining the corresponding model input feature by the current service information including the attribute information of the current customer service order is further described by the following third embodiment:

EXAMPLE III

Here, in the third embodiment of the present application, it is considered that the attribute information of whether the work order is a work order with repeated incoming lines may reflect whether the driver/passenger may affect the incoming line because of previous incoming line dissatisfaction, and the IVR waiting duration may reflect the incoming line experience of the driver/passenger, that is, it is important that the two user satisfaction affecting factors are taken as relevant factors directly affecting the specific situation of the work order. Therefore, in a specific implementation, the above-mentioned incoming repeated work order and IVR waiting time may be subjected to a characterization process, and the corresponding characteristic value may be used as a model input characteristic.

In addition, in the embodiment of the application, other attribute information of the current customer service work order, such as the work order state, the processing person ID, the processing department ID, the ringing duration, the waiting duration and other related information, can be selected to perform feature value processing in combination with a specific application scene, so as to obtain the corresponding model input feature. The embodiment of the application does not specifically limit the selection of the attribute information of the current customer service work order so as to meet the requirements of various application scenes.

The process of determining the corresponding model input features by the current service information including the current order information is further explained by the following fourth embodiment:

example four

Here, in the fourth embodiment of the present application, it is considered that whether the current order information is automatically responsible for determining may reflect the incoming line problem of the driver/passenger and the possibility of whether the current order information is satisfied, and the waiting time of the car booking user may reflect the car using experience of the passenger, that is, the two user satisfaction influence factors are important as the relevant factors directly influencing the specific situation of the current order. Therefore, in the specific implementation, the automatic responsibility judgment and the waiting time of the network appointment user can be subjected to characteristic value processing, and the corresponding characteristic value can be used as the model input characteristic.

In addition, in the embodiment of the application, other current order information, such as information about estimated distance, whether detour is performed, pre-estimated value, waiting time, driving receiving distance, issuing and responding time difference, order cancellation type, responsibility judgment result, whether peak is performed in the morning and the evening, and the like, can be selected to perform characteristic value processing in combination with a specific application scene, so as to obtain corresponding model input characteristics. The embodiment of the application does not specifically limit the selection of the current order information to meet the requirements of various application scenarios.

In the embodiment of the application, the training process of the user satisfaction prediction model is a key step of user satisfaction prediction. The training process of the user satisfaction prediction model is described in the following fifth embodiment.

EXAMPLE five

As shown in fig. 3, the training process of the user satisfaction prediction model is specifically implemented by the following steps:

s301, obtaining user attribute information of each sample user, historical service information provided by a platform for the sample user, customer service attribute information of customer service serving the sample user, and satisfaction degree scoring information of the sample user for the customer service;

s302, taking the user attribute information, the historical service information and the customer service attribute information as user satisfaction influence factors, taking the satisfaction scoring information as a user satisfaction result, and training to obtain a user satisfaction prediction model.

Here, training of the user satisfaction prediction model may be performed based on the user attribute information, the historical service information, the customer service attribute information, and the satisfaction scoring information obtained as described above in relation to the sample user. In the stage of training the user satisfaction prediction model, the user attribute information, the customer service attribute information and the historical service information acquired in the step S301 are used as user satisfaction influence factors, and the satisfaction scoring information is used as a user satisfaction result, so that the training parameters of the user satisfaction prediction model can be obtained through training, that is, the trained user satisfaction prediction model is obtained. In the embodiment of the application, an integrated tree model, such as an Extreme Gradient Boosting tree (XGBOOST) model, may be used as the user satisfaction prediction model, and the model training phase is a process of training some unknown parameter information in the model.

In order to facilitate training of the user satisfaction prediction model according to the user satisfaction influence factors, the user satisfaction influence factors can be subjected to characteristic value processing before model training, that is, in the embodiment of the application, the user satisfaction influence factors can be subjected to characteristic value processing firstly, the model input characteristics corresponding to the sample user are determined, then the model input characteristics are used as independent variables, and the satisfaction scoring information is used as dependent variables to perform at least one round of model training to obtain the user satisfaction prediction model.

The user attribute information of the sample user may be information corresponding to the user attribute information of the target user, and the customer service attribute information of the customer service of the service sample user may be information corresponding to the customer service attribute information of the customer service of the service target user, which is not described herein again. The historical service information may only include historical order information generated by historical order taking of the whole network car booking travel platform, may only include relevant information of a historical customer service work order generated when a user enters a line to a customer service according to whether the order is satisfied or not after the order is completed (such as attribute information describing the specific situation of the work order and recording content when the customer service communicates with the user for specific problems), and may include both the historical order information and the relevant information of the historical customer service work order. In consideration of the fact that in an actual application scenario, a travel order and a customer service work order often have close correlation, in the embodiment of the present application, the integrated information of the historical order information and the information related to the historical customer service work order may be used as the historical service information.

In addition, in the embodiment of the application, the satisfaction degree scoring information of the driver/passenger on the incoming line customer service can be used as one measurement for measuring the satisfaction degree of the user on the customer service. In the evaluation of the incoming line customer service by the driver/passenger, a score of 1 to 6 points is included, wherein 4 points or more can be used as a satisfactory customer service order, and less than 4 points can be used as an unsatisfactory customer service order, and the good or bad of the incoming line satisfaction degree is represented by increasing or decreasing the score, namely, the satisfaction degree of the driver/passenger on the incoming line customer service is improved along with the increase of the satisfaction degree score and is reduced along with the reduction of the satisfaction degree score. Therefore, the user satisfaction degree prediction model can be simplified into a two-classification model, and the method is simple and easy to implement and high in applicability.

On the basis of the fifth embodiment, whether the target user is satisfied with the customer service serving the target user can be determined according to the comparison result between the determined user satisfaction prediction result and the preset satisfaction threshold, and when the target user is not satisfied, customer dissatisfaction prompt information is pushed to a responsible person. If 4 points corresponding to the fifth embodiment are used as the preset satisfaction threshold, when the user satisfaction prediction result obtained through the prediction of the user satisfaction prediction model is greater than or equal to 4, the customer service satisfaction of the target user on the service is determined, and if the user satisfaction prediction result obtained through the prediction is less than 4, the customer service satisfaction of the target user on the service is determined.

In order to further improve the pertinence of satisfaction degree prediction, when the target user is determined to be unsatisfied with the customer service of the service, on one hand, a keyword matching method of a work order record and a service dictionary can be used for directly capturing the cause of pain points in a targeted manner, namely determining the cause of dissatisfaction with the customer service; on the other hand, whether the latest historical order information of the user meets the established rule can be traced, namely, the work order is attributed and classified by the rule matching method based on the order information. Next, the above two aspects will be specifically explained by the following examples six and seven.

Firstly, the procedure of dissatisfaction type classification by the keyword matching method based on the work order record and the service dictionary is explained through the sixth embodiment.

EXAMPLE six

Here, in the sixth embodiment of the present application, the obtained current service information provided by the platform for the target user may be subjected to feature value processing, and then the processed current service information is input into a pre-trained dissatisfaction classification model, so that a dissatisfaction type prediction result of the target user for customer service may be obtained. That is, the dissatisfaction type of the target user to the customer service can be predicted through the pre-trained dissatisfaction classification model in the embodiment of the application. The characteristic value processing procedure of the current service information for the dissatisfaction classification model may be the same as or different from the characteristic value processing procedure of the current service information for the user satisfaction prediction model, and this is not specifically limited in the embodiments of the present application.

As shown in fig. 4, the training process of the unsatisfactory classification model specifically includes the following steps:

s401, acquiring the record content of the historical customer service work order corresponding to each sample user;

s402, extracting dissatisfaction label classification information of the sample user from the acquired record content of the historical customer service work order;

and S403, taking the recorded content of the historical customer service order as an influence factor of unsatisfactory classification, taking the unsatisfactory label classification information as an unsatisfactory classification result, and training to obtain the unsatisfactory classification model.

Here, similar to the user satisfaction prediction model, the embodiment of the present application needs to analyze and process the related information of the sample user, and different from the user satisfaction prediction model, the user satisfaction prediction model is limited by the influence of each user satisfaction influence factor, and needs user attribute information, customer service attribute information, historical service information, and the like, and for training of the dissatisfaction classification model, the training mainly depends on the record content of the historical customer service work order in the historical service information, which mainly considers the customer service incoming line as a trigger point of problem development, and the problem recording is an important basis. Based on the above content, the embodiment of the application can count the keywords of each type of problem aiming at the problems pointed out in the customer service work orders of the historical incoming lines, take the counted keywords as the dissatisfaction label classification information of the sample user, take the dissatisfaction label classification information as the dissatisfaction classification result, and take the recorded content of the corresponding historical customer service work orders as the influence factors of the dissatisfaction classification to train the dissatisfaction classification model. The unsatisfied classification model can be specifically realized based on a multi-pattern matching Aho-Corasick algorithm.

Next, the procedure of dissatisfaction type classification by the rule matching method based on order information will be described through a seventh embodiment.

EXAMPLE seven

As shown in fig. 5, the determination process of the dissatisfaction cause category specifically includes the following steps:

s501, determining the latest historical order information of the target user corresponding to the customer service work order;

s502, aiming at multiple preset dissatisfaction reason categories, extracting order information concerned by each dissatisfaction reason category from the determined historical order information;

s503, comparing the extracted order information concerned by each dissatisfaction reason category with dissatisfaction order information corresponding to the dissatisfaction reason category, and determining whether the customer service order conforms to the dissatisfaction reason category.

Here, considering that a part of the customer service work order is associated with the previous order of the user, for example, the user may make an incoming consultation due to a problem that the difference between the estimated price of the previous order and the actual payment price is large, the last historical order information corresponding to the customer service work order is traced back on the basis of the recorded content of the customer service work order, so that it is ensured that the specific dissatisfaction reason information is not lost. In this way, the order information concerned by each dissatisfaction reason category is extracted from the determined historical order information according to various preset dissatisfaction reason categories, and whether the customer service order conforms to the dissatisfaction reason category is determined by comparing the extracted order information concerned by each dissatisfaction reason category with the dissatisfaction order information corresponding to the dissatisfaction reason category. That is, according to the embodiment of the application, the order information concerned by each dissatisfaction reason category in the multiple preset dissatisfaction reason categories can be extracted from the latest historical order information of the target user, the dissatisfaction reason category met by the current customer service work order is determined according to the comparison result of the order information, the pertinence of the satisfaction prediction is further ensured, and the applicability is stronger.

Based on the first to seventh embodiments, the present application further provides a user satisfaction prediction apparatus, and the following various apparatuses may be implemented by referring to the method, and repeated details are not repeated.

Example eight

As shown in fig. 6, a schematic structural diagram of a user satisfaction predicting apparatus provided in the ninth embodiment of the present application includes:

an information obtaining module 601, configured to obtain user attribute information of a target user, current service information provided by a platform for the target user, and customer service attribute information of customer service serving the target user;

a feature determination module 602, configured to perform feature value processing on the obtained user attribute information of the target user, current service information provided by the platform for the target user, and customer service attribute information of customer service serving the target user, and determine a model input feature corresponding to the target user;

and the satisfaction predicting module 603 is configured to input the determined model input features into a pre-trained user satisfaction predicting model to obtain a user satisfaction predicting result of the target user.

In some embodiments, the current service information includes recorded content of a current customer service order; the feature determining module 602 is specifically configured to:

the feature determining module 602 is specifically configured to:

In one embodiment, the apparatus further comprises:

the determining module 604 is configured to compare the determined user satisfaction prediction result with a preset satisfaction threshold, determine whether the target user is satisfied with the customer service serving the target user, and if the target user is not satisfied with the customer service, push a customer dissatisfaction prompt message to a responsible person.

an unsatisfied category determining module 605, configured to determine the latest historical order information of the target user corresponding to the customer service work order;

In yet another embodiment, the apparatus further comprises:

and an unsatisfied type prediction module 606, configured to perform eigenvalue processing on the obtained current service information provided by the platform for the target user, and input the processed eigenvalue into a pre-trained unsatisfied classification model to obtain an unsatisfied type prediction result of the target user for customer service.

In yet another embodiment, the apparatus further comprises:

the classification model training module 607 is configured to obtain the record content of the historical customer service order corresponding to each sample user;

In some embodiments, the apparatus further comprises:

the prediction model training module 608 is configured to obtain user attribute information of each sample user, historical service information provided by the platform for the sample user, customer service attribute information of customer service serving the sample user, and satisfaction scoring information of the sample user for the customer service;

In some embodiments, the predictive model training module 608 is specifically configured to:

Example nine

As shown in fig. 7, a schematic structural diagram of an electronic device provided in a ninth embodiment of the present application includes: a processor 701, a storage medium 702 and a bus 703, wherein the storage medium 702 stores machine-readable instructions executable by the processor 701, when the electronic device is operated, the processor 701 and the storage medium 702 communicate via the bus 703, and the machine-readable instructions, when executed by the processor 701, perform the following:

In a specific implementation, the current service information comprises the record content of the current customer service work order; in the processing executed by the processor 701, the determining the model input feature corresponding to the target user includes:

In a specific implementation, the current service information includes attribute information of a current customer service work order; wherein, the attribute information of the current customer service work order comprises: whether the waiting time is the work order of repeated incoming lines and the interactive voice response IVR waiting time; in the processing executed by the processor 701, the determining the model input feature corresponding to the target user includes:

In a specific implementation, the current service information includes current order information; wherein the current order information comprises: whether to automatically judge responsibility and the waiting time of the user; in the processing executed by the processor 701, the determining the model input feature corresponding to the target user includes:

In one embodiment, the processing performed by the processor 701 further includes:

In another embodiment, the current service information provided by the platform for the target user includes the recorded content of the customer service work order; when it is determined that the target user is not satisfied with the service of the service, the processing performed by the processor 701 further includes:

In another embodiment, when it is determined that the target user is not satisfied with the customer service of the target user, the processing performed by the processor 701 further includes:

In a specific implementation, in the processing performed by the processor 701, the unsatisfactory classification model is trained according to the following steps:

In a specific implementation, in the processing executed by the processor 701, the user satisfaction prediction model is trained according to the following steps:

In another embodiment, in the processing executed by the processor 701, the training to obtain the user satisfaction prediction model by using the user attribute information, the historical service information, and the customer service attribute information as user satisfaction influencing factors, and using the satisfaction score information as a user satisfaction result includes:

Example ten

An embodiment tenth of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor 701, the method for predicting user satisfaction according to any of the first to seventh embodiments is performed.

Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, and the like, and when a computer program on the storage medium is run, the user satisfaction prediction method can be executed, so that the problems that the evaluation efficiency is low in the current investigation mode, the investigation recovery amount is small, and the user pain point cannot be known in time due to the fact that the investigation recovery amount is not statistical, and the effects of prediction of the satisfaction of any user and high predicted coverage rate and efficiency are achieved.

The computer program product of the user satisfaction prediction method provided in the embodiment of the present application includes a computer-readable storage medium storing a program code, where instructions included in the program code may be used to execute the method in the foregoing method embodiment, and specific implementation may refer to the method embodiment, which is not described herein again.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the system and the apparatus described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A user satisfaction prediction method, comprising:

2. The method of claim 1, wherein the current service information includes recorded content of a current customer service order; the determining of the model input features corresponding to the target user includes:

3. The method of claim 1, wherein the current service information includes attribute information of a current customer service order; wherein, the attribute information of the current customer service work order comprises: whether the waiting time is the work order of repeated incoming lines and the interactive voice response IVR waiting time;

4. The method of claim 1, wherein the current service information comprises current order information; wherein the current order information comprises: whether to automatically judge responsibility and the waiting time of the user;

5. The method of claim 1, further comprising:

6. The method of claim 5, wherein the current service information provided by the platform for the target user includes recorded content of a customer service order; when the target user is determined to be not satisfied with the customer service of the service, the method further comprises the following steps:

7. The method of claim 5, wherein upon determining that the target user is not satisfied with customer service to service the target user, further comprising:

8. The method of claim 7, wherein said unsatisfactory classification model is trained according to the steps of:

9. The method of any of claims 1 to 8, wherein the user satisfaction prediction model is trained according to the following steps:

10. The method of claim 9, wherein training a user satisfaction prediction model using the user attribute information, the historical service information, and the customer service attribute information as user satisfaction influencing factors and the satisfaction scoring information as a user satisfaction result comprises:

11. A user satisfaction prediction apparatus, comprising:

12. The apparatus of claim 11, wherein the current service information comprises recorded content of a current customer service order; the feature determination module is specifically configured to:

13. The apparatus of claim 11, wherein the current service information comprises attribute information of a current customer service order; wherein, the attribute information of the current customer service work order comprises: whether the waiting time is the work order of repeated incoming lines and the interactive voice response IVR waiting time;

the feature determination module is specifically configured to:

14. The apparatus of claim 11, wherein the current service information comprises current order information; wherein the current order information comprises: whether to automatically judge responsibility and the waiting time of the user;

the feature determination module is specifically configured to:

15. The apparatus of claim 11, further comprising:

16. The apparatus of claim 15, wherein the current service information provided by the platform for the target user comprises recorded content of a customer service order; further comprising:

17. The apparatus of claim 15, further comprising:

18. The apparatus of claim 17, further comprising:

19. The apparatus of any one of claims 11 to 18, further comprising:

20. The apparatus of claim 19, wherein the predictive model training module is specifically configured to:

21. An electronic device, comprising: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor and the storage medium communicating via the bus when the electronic device is running, the processor executing the machine-readable instructions to perform the steps of the user satisfaction prediction method according to any of claims 1 to 10.

22. A computer-readable storage medium, having stored thereon a computer program for performing, when being executed by a processor, the steps of the user satisfaction prediction method according to any of the claims 1 to 10.