Target impact factor determination method and device and storage medium

Abstract

The invention relates to a target impact factor determination method and device and a storage medium, and the method comprises the steps: obtaining a plurality of pieces of actual data corresponding to a plurality of actual users, carrying out the discretization of impact factor satisfaction degrees of a plurality of impact factors of the plurality of actual users, obtaining a target group of each actual user about each impact factor, processing the target group of each actual user about each influence factor according to a score mapping model to obtain an influence factor score of each actual user about each influence factor, and determining a migration score of each influence factor according to the influence factor score of each actual user about each influence factor; and finally, determining the influence factor corresponding to the maximum migration score as a target influence factor. According to the embodiment of the invention, the target influence factor with the maximum improvement on the overall satisfaction can be determined, so that service adjustment measures can be implemented according to the target influence factor, and the overall satisfaction of users can be improved to the maximum extent.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a method, device and storage medium for determining a target impact factor. Background Art

[0002] As for the overall satisfaction with the service, we can obtain the evaluation of the overall satisfaction of different users on the service so as to improve the service and enhance the overall satisfaction with the service.

[0003] In the related art, there are many factors affecting the overall satisfaction with the service. Since the various factors have different impacts on the overall satisfaction with the service, it is impossible to determine more effective factors to improve the overall satisfaction. Summary of the invention

[0004] In order to overcome the problems existing in the related art, the present disclosure provides a method, device and storage medium for determining a target influencing factor, by obtaining multiple actual data corresponding to multiple actual users, and determining the target grouping of each actual user with respect to each influencing factor based on the multiple actual data, and then training the basic network with multiple sample overall satisfaction and multiple sample influencing factor satisfaction corresponding to multiple sample users to obtain an overall satisfaction prediction model, and constructing a score mapping model based on the overall satisfaction prediction model, so as to process the target grouping of each actual user with respect to each influencing factor according to the score mapping model, obtain the influencing factor score of each actual user with respect to each influencing factor, and further determine the migration score of each influencing factor, so as to determine the target influencing factor with the largest migration score. Through the above method, the quantitative relationship between each influencing factor and the overall satisfaction can be reflected through the influencing factor score of each actual user with respect to each influencing factor, so as to determine the target influencing factor that can improve the overall satisfaction the most, so as to implement business adjustment measures according to the target influencing factor, so as to maximize the overall satisfaction of the user.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for determining a target impact factor is provided, including:

[0006] Acquire a plurality of actual data corresponding to a plurality of actual users, the actual data including actual overall satisfaction and actual influencing factor satisfaction of a plurality of influencing factors associated with the actual overall satisfaction;

[0007] Discretize the actual influencing factor satisfaction of the multiple influencing factors of the multiple actual users to obtain a target grouping of each actual user with respect to each influencing factor;

[0008] The target grouping of each actual user with respect to each influencing factor is processed respectively according to the score mapping model to obtain the influencing factor score of each actual user with respect to each influencing factor, wherein the score mapping model is constructed based on the overall satisfaction prediction model, and the overall satisfaction prediction model is obtained by training a basic network based on multiple sample overall satisfactions and multiple sample influencing factor satisfactions corresponding to multiple sample users;

[0009] Determine a migration score for each influencing factor based on the influencing factor score of each actual user for each influencing factor, wherein the migration score represents the degree to which the influencing factor improves overall satisfaction;

[0010] The impact factor corresponding to the maximum migration score is determined as the target impact factor.

[0011] Optionally, the overall satisfaction prediction model is trained by the following steps:

[0012] Acquire a plurality of sample data corresponding to a plurality of sample users, wherein the sample data includes sample overall satisfaction and sample influencing factor satisfaction of a plurality of influencing factors associated with the sample overall satisfaction;

[0013] Determine an overall satisfaction label for each sample user based on a preset overall satisfaction threshold and the overall satisfaction of the sample;

[0014] Discretizing the influence factor satisfaction of the multiple influence factors of the multiple sample users to obtain an original sample grouping of each sample user with respect to each influence factor;

[0015] According to the overall satisfaction label of each sample user, the evidence weight value of each original sample group is obtained;

[0016] According to the evidence weight value of each sample user's original sample grouping with respect to the multiple influencing factors and the overall satisfaction label of each sample user, the basic network is trained for multiple rounds to obtain the overall satisfaction prediction model.

[0017] Optionally, before performing multiple rounds of training on the basic network according to the evidence weight values ​​of the original sample groupings of each sample user with respect to the multiple influencing factors and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model, the method further includes:

[0018] According to the evidence weight value of each original sample group, the multiple influencing factors are screened to obtain initial influencing factors and multiple updated sample groups corresponding to each initial influencing factor;

[0019] According to the overall satisfaction label of each sample user, the evidence weight value of each updated sample group is obtained;

[0020] The method of training the basic network for multiple rounds according to the evidence weight values ​​of the original sample groupings of each sample user with respect to the multiple influencing factors and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model includes:

[0021] According to the evidence weight value of each sample user's updated sample grouping with respect to the initial influencing factor and the overall satisfaction label of each sample user, the basic network is trained for multiple rounds to obtain the overall satisfaction prediction model.

[0022] Optionally, the screening of the multiple influencing factors according to the evidence weight value of each original sample group to obtain the initial influencing factors and the multiple updated sample groups corresponding to each initial influencing factor includes:

[0023] For any influencing factor among the multiple influencing factors, according to the multiple original sample groups corresponding to the any influencing factor, determine the actual monotonic relationship of the evidence weight values ​​of the multiple original sample groups;

[0024] In the case where the actual monotonic relationship is monotonically increasing, any one of the influencing factors is determined to be an initial influencing factor, and the sample groupings in the monotonically increasing state are determined to be a plurality of updated sample groups corresponding to the initial influencing factors.

[0025] Optionally, for any influencing factor among the multiple influencing factors, determining the actual monotonic relationship of the evidence weight values ​​of the multiple original sample groups according to the multiple original sample groups corresponding to the any influencing factor includes:

[0026] Determine a plurality of original sample groups corresponding to any of the influencing factors as a plurality of basic groups;

[0027] Sorting the multiple basic groups and determining a sequence number for each basic group;

[0028] When the monotonicity of the evidence weight values ​​of the plurality of basic groups after sorting is non-monotonic, determining a target sequence number of the basic group corresponding to the non-monotonic target evidence weight value;

[0029] According to the target sequence number, the multiple basic groups are combined to obtain a first group sequence;

[0030] In the case where the monotonicity of the evidence weight values ​​of the first grouping sequence is not monotonic, a plurality of new basic groups are determined according to the first grouping sequence, and the evidence weight value of each new basic group is determined until the number of the plurality of new basic groups is less than the preset number of groups, or the evidence weight values ​​of the plurality of new basic groups are monotonic;

[0031] The basic grouping when the evidence weight value is monotonic is determined as a plurality of updated sample groups;

[0032] The monotonic relationship among the evidence weight values ​​of the multiple updated sample groups is determined as the actual monotonic relationship among the evidence weight values ​​of the multiple original sample groups.

[0033] Optionally, the merging the multiple basic groups according to the target sequence number to obtain a first group sequence includes:

[0034] In the case where the target sequence number is the first sequence number, based on the group corresponding to the target sequence number, basic groups of adjacent sequence numbers are merged backward to obtain the first group sequence.

[0035] Optionally, the merging the multiple basic groups according to the target sequence number to obtain a first group sequence includes:

[0036] In the case where the target sequence number is the last sequence number, based on the group corresponding to the target sequence number, basic groups of adjacent sequence numbers are merged forward to obtain the first group sequence.

[0037] Optionally, the first packet sequence includes a first packet subsequence and a second packet subsequence;

[0038] The step of merging the plurality of basic groups according to the target sequence number to obtain a first group sequence includes:

[0039] When the target sequence number is an intermediate sequence number, the basic groups of adjacent sequence numbers are merged backward based on the group corresponding to the target sequence number to obtain a first group subsequence, and the basic groups of adjacent sequence numbers are merged forward based on the group corresponding to the target sequence number to obtain a second group subsequence.

[0040] Optionally, when the monotonicity of the evidence weight values ​​of the first grouping sequence is non-monotonic, determining a plurality of new basic groups according to the first grouping sequence includes:

[0041] In a case where the monotonicity of the evidence weight values ​​of the first grouping subsequence and the second grouping subsequence are both non-monotonic, determining a first information value of an influencing factor corresponding to the first grouping subsequence, and determining a second information value of an influencing factor corresponding to the second grouping subsequence;

[0042] The group subsequences corresponding to the larger value of the first information value and the second information value are determined as a plurality of new basic groups.

[0043] Optionally, before performing multiple rounds of training on the basic network according to the evidence weight value of the updated sample grouping of each sample user with respect to the initial influencing factor and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model, the method further includes:

[0044] Determine the third information value of each initial impact factor;

[0045] Determine the initial impact factor whose third information value is greater than the preset information value threshold as the first target impact factor;

[0046] Determine the correlation coefficient between any two first target impact factors according to the evidence weight value of each updated sample group corresponding to each first target impact factor;

[0047] Determine a screening influence factor according to the correlation coefficient, wherein the screening influence factor is the first target influence factor with the smaller third information value among the two first target influence factors whose correlation coefficient is greater than a preset correlation coefficient threshold;

[0048] Deleting the screening impact factor from the first target impact factor to obtain a second target impact factor;

[0049] The method of performing multiple rounds of training on the basic network according to the evidence weight value of the updated sample grouping of each sample user with respect to the initial influencing factor and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model includes:

[0050] According to the evidence weight value of the updated sample grouping of each sample user regarding the second target influencing factor and the overall satisfaction label of each sample user, the basic network is trained for multiple rounds to obtain the overall satisfaction prediction model.

[0051] Optionally, determining the migration score of each influencing factor according to the influencing factor score of each actual user with respect to each influencing factor includes:

[0052] Determine, according to the influence factor score of each actual user with respect to each influence factor, the influence factor score corresponding to each group of each influence factor and the number of actual users corresponding to each group;

[0053] Determine a migration group and a target group for each impact factor according to the impact factor score corresponding to each group of each impact factor, wherein the impact factor score corresponding to the target group is greater than the impact factor score corresponding to the migration group;

[0054] The migration score of each influencing factor is determined according to the influencing factor score corresponding to the migration group of each influencing factor, the influencing factor score corresponding to the target group, and the number of actual users corresponding to the migration group.

[0055] According to a second aspect of an embodiment of the present disclosure, there is provided a device for determining a target impact factor, including:

[0056] A first acquisition module is configured to acquire a plurality of actual data corresponding to a plurality of actual users, wherein the actual data includes an actual overall satisfaction degree and actual influencing factor satisfaction degrees of a plurality of influencing factors associated with the actual overall satisfaction degree;

[0057] A discretization module is configured to discretize the actual influencing factor satisfaction of the multiple influencing factors of the multiple actual users to obtain a target grouping of each actual user with respect to each influencing factor;

[0058] A first obtaining module is configured to process the target grouping of each actual user with respect to each influencing factor respectively according to a score mapping model to obtain an influencing factor score of each actual user with respect to each influencing factor, wherein the score mapping model is constructed based on an overall satisfaction prediction model, and the overall satisfaction prediction model is obtained by training a basic network based on a plurality of sample overall satisfactions and a plurality of sample influencing factor satisfactions corresponding to a plurality of sample users;

[0059] A first determination module is configured to determine a migration score of each influencing factor according to the influencing factor score of each actual user with respect to each influencing factor, wherein the migration score represents the degree to which the influencing factor improves the overall satisfaction;

[0060] The second determination module is configured to determine the impact factor corresponding to the maximum migration score as the target impact factor.

[0061] According to a third aspect of an embodiment of the present disclosure, there is provided a device for determining a target impact factor, including:

[0062] processor;

[0063] a memory for storing processor-executable instructions;

[0064] Among them, the processor is configured to implement the steps of the target impact factor determination method provided in the first aspect of the present disclosure.

[0065] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the program instructions are executed by a processor, the steps of the target influence factor determination method provided in the first aspect of the present disclosure are implemented.

[0066] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0067] By obtaining multiple actual data corresponding to multiple actual users, the actual data includes actual overall satisfaction and actual influencing factor satisfaction of multiple influencing factors associated with the actual overall satisfaction, and discretizing the influencing factor satisfaction of multiple influencing factors of multiple actual users, obtaining the target grouping of each actual user with respect to each influencing factor, and then processing the target grouping of each actual user with respect to each influencing factor according to the score mapping model, to obtain the influencing factor score of each actual user with respect to each influencing factor, the score mapping model is constructed based on the overall satisfaction prediction model, and the overall satisfaction prediction model is obtained by training a basic network based on multiple sample overall satisfactions and multiple sample influencing factor satisfactions corresponding to multiple sample users, and according to the influencing factor score of each actual user with respect to each influencing factor, the migration score of each influencing factor is determined, and the migration score represents the degree to which the influencing factor improves the overall satisfaction, and finally the influencing factor corresponding to the maximum migration score is determined as the target influencing factor.

[0068] By obtaining multiple actual data corresponding to multiple actual users, and determining the target grouping of each actual user for each influencing factor based on the multiple actual data, and then training the basic network through multiple sample overall satisfaction and multiple sample influencing factor satisfaction corresponding to multiple sample users to obtain an overall satisfaction prediction model, and constructing a score mapping model based on the overall satisfaction prediction model, so as to process the target grouping of each actual user for each influencing factor according to the score mapping model, obtain the influencing factor score of each actual user for each influencing factor, and further determine the migration score of each influencing factor, so as to determine the target influencing factor with the largest migration score. Through the above method, the quantitative relationship between each influencing factor and the overall satisfaction can be reflected through the influencing factor score of each actual user for each influencing factor, so as to determine the target influencing factor that can improve the overall satisfaction the most, so as to implement business adjustment measures according to the target influencing factor, so as to maximize the overall satisfaction of users.

[0069] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0071] Figure 1 The figure is a flow chart of a method for determining a target impact factor according to an exemplary embodiment.

[0072] Figure 2 The present invention is a flowchart of a method for training an overall satisfaction prediction model according to an exemplary embodiment.

[0073] Figure 3 The figure is a flow chart of a method for screening influencing factors according to an exemplary embodiment.

[0074] Figure 4 The invention is a flow chart showing a method for determining an actual monotonic relationship according to an exemplary embodiment.

[0075] Figure 5 The invention is a flowchart of a method for determining a second target impact factor according to an exemplary embodiment.

[0076] Figure 6 The figure is a flow chart showing a method for determining a migration score according to an exemplary embodiment.

[0077] Figure 7 The present invention is a block diagram of a device for determining a target impact factor according to an exemplary embodiment.

[0078] Figure 8 The figure is a block diagram of another device for determining a target impact factor according to an exemplary embodiment. DETAILED DESCRIPTION

[0079] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0080] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the relevant data protection laws and policies of the country where the device is located and with the authorization given by the owner of the corresponding device.

[0081] For the overall satisfaction of services, such as sales services, we can obtain different users' evaluations of their overall satisfaction with the services so as to improve the services and enhance the overall satisfaction with the services.

[0082] In the related art, there are multiple factors affecting the overall satisfaction with the service. Since multiple factors have different effects on the overall satisfaction with the service, for example, the overall satisfaction may be the overall satisfaction with the shopping service, and the multiple factors affecting the overall satisfaction may be logistics timeliness and delivery services, etc. Different factors have different effects on improving the overall satisfaction of users, and it is impossible to determine a more effective factor from multiple factors to improve the overall satisfaction.

[0083] Based on the above technical problems, the present disclosure provides a method, device and storage medium for determining a target influencing factor, by obtaining multiple actual data corresponding to multiple actual users, and determining the target grouping of each actual user with respect to each influencing factor based on the multiple actual data, and then training the basic network with multiple sample overall satisfaction and multiple sample influencing factor satisfaction corresponding to multiple sample users to obtain an overall satisfaction prediction model, and constructing a score mapping model based on the overall satisfaction prediction model, so as to process the target grouping of each actual user with respect to each influencing factor according to the score mapping model, obtain the influencing factor score of each actual user with respect to each influencing factor, and further determine the migration score of each influencing factor, so as to determine the target influencing factor with the largest migration score. Through the above method, the quantitative relationship between each influencing factor and the overall satisfaction can be reflected through the influencing factor score of each actual user with respect to each influencing factor, so as to determine the target influencing factor that can improve the overall satisfaction the most, so as to implement business adjustment measures according to the target influencing factor, so as to maximize the overall satisfaction of the user.

[0084] For example, for shopping services, multiple sample data corresponding to multiple sample users can be obtained, and the sample data can include shopping satisfaction data of multiple sample users. The shopping satisfaction data may include the sample overall satisfaction and multiple sample influencing factor satisfactions corresponding to multiple influencing factors. An overall satisfaction prediction model is trained based on the multiple sample data, and a score mapping model is constructed through the overall satisfaction prediction model. The score mapping model characterizes the mapping relationship between the target grouping of each influencing factor and the influencing factor score. The target grouping of each actual user regarding each influencing factor is processed through the score mapping model to obtain the influencing factor score of each actual user regarding each influencing factor, and further determine the migration score of each influencing factor, so as to determine the target influencing factor with the largest migration score.

[0085] Figure 1 is a flow chart of a method for determining a target impact factor according to an exemplary embodiment. Figure 1 As shown, the method can be applied to a server and includes the following steps.

[0086] In step S101, a plurality of actual data corresponding to a plurality of actual users are obtained, wherein the actual data include actual overall satisfaction and actual influencing factor satisfaction of a plurality of influencing factors associated with the actual overall satisfaction.

[0087] In this embodiment, the multiple actual users may be users related to the target service, and the target service may be a sales service, an online shopping service, and the like. For example, for the online shopping service, the multiple actual users may be multiple shopping users in the recent period of time. The multiple actual data may be obtained through a questionnaire survey, for example, a satisfaction questionnaire is sent to multiple actual users, and the satisfaction questionnaire may be a total score structure, and the satisfaction questionnaire may inquire about the overall satisfaction and the influencing factor satisfaction of multiple influencing factors associated with the overall satisfaction, such as the logistics progress inquiry satisfaction and the logistics timeliness satisfaction, etc. The overall satisfaction and the influencing factor satisfaction are both interval values, such as integer values ​​between 1 and 10. The overall satisfaction represents the user's overall satisfaction with the target service, and the higher the value, the greater the satisfaction; the influencing factor satisfaction represents the user's satisfaction with a single or multiple influencing factors of the target service, and the higher the value, the greater the satisfaction.

[0088] Influencing factors can affect the overall satisfaction of users. For example, influencing factors may be shipping warehouse, carrier, shipping place, receiving place, purchase channel, delivery time, and delivery time. Overall satisfaction can be associated with multiple influencing factors. Through the satisfaction survey form, the actual overall satisfaction and multiple actual influencing factor satisfaction corresponding to each actual user can be obtained.

[0089] In step S102, the actual influencing factor satisfaction levels of multiple influencing factors of multiple actual users are discretized to obtain a target grouping of each actual user with respect to each influencing factor.

[0090] In this embodiment, the actual influence factor satisfaction of multiple influence factors can be discretized based on the grouping parameters of the actual influence factor satisfaction of each influence factor when constructing the score mapping model, and the target grouping of each actual user for each influence factor can be obtained to facilitate the processing of the score mapping model. The target grouping represents the grouping number of the actual user for the influence factor and carries the influence factor information, that is, which influence factor the target grouping belongs to. For the actual influence factor satisfaction, the grouping parameter can be at least one grouping threshold, for example, the grouping threshold can be 2, 6 and 9, and the actual influence factor satisfaction is between 0 and 2 as a group, between 2 and 6 as a group, between 6 and 9 as a group, and between 9 and 10 as a group.

[0091] In step S103, the target grouping of each actual user regarding each influencing factor is processed respectively according to the score mapping model to obtain the influencing factor score of each actual user regarding each influencing factor. The score mapping model is constructed based on the overall satisfaction prediction model, and the overall satisfaction prediction model is obtained by training a basic network based on multiple sample overall satisfactions and multiple sample influencing factor satisfactions corresponding to multiple sample users.

[0092] In this embodiment, multiple sample users may be users related to the target service, and the target service may be a sales service, an online shopping service, and the like. For example, for online shopping services, multiple sample users may be multiple shopping users in the recent period of time. Multiple sample data corresponding to multiple sample users may be obtained, and the multiple sample data may include multiple sample overall satisfaction and multiple sample influencing factor satisfaction corresponding to multiple sample users. The multiple sample data may be obtained through a questionnaire survey, for example, a satisfaction survey form may be sent to multiple sample users, and the satisfaction survey form may be a total score structure, and the satisfaction survey form may inquire about the overall satisfaction and the influencing factor satisfaction of multiple influencing factors associated with the overall satisfaction, such as logistics progress inquiry satisfaction and logistics timeliness satisfaction, etc. The overall satisfaction and influencing factor satisfaction are both interval values, such as integer values ​​between 1 and 10. One group corresponds to one influencing factor score, and multiple actual users with the same target group have the same influencing factor score for the influencing factor.

[0093] Influencing factors can affect the overall satisfaction of users. For example, influencing factors may be shipping warehouse, carrier, shipping place, receiving place, purchase channel, delivery time, and delivery time. Overall satisfaction can be associated with multiple influencing factors. Through the satisfaction survey form, the overall satisfaction of each sample user and the satisfaction of multiple sample influencing factors can be obtained.

[0094] Then, the basic network is trained by using multiple sample overall satisfaction and multiple sample influencing factor satisfaction corresponding to multiple sample users to obtain an overall satisfaction prediction model, wherein the basic network can be a basic Logistic regression model. Based on the overall satisfaction prediction model, a score mapping model is constructed according to the target mapping interval and the target mapping rule, wherein the target mapping rule represents the linear mapping relationship between overall satisfaction and overall satisfaction score, and the target mapping interval is the numerical interval of the overall satisfaction score, which can be a pre-set interval. The target mapping interval is determined by the maximum score value and the minimum score value. For example, the minimum score value min_score can be 0, and the maximum score value max_score can be 10, then the target mapping interval can be [0,10].

[0095] In step S104, a migration score of each influencing factor is determined according to the influencing factor score of each actual user with respect to each influencing factor, and the migration score represents the degree to which the influencing factor improves the overall satisfaction.

[0096] In this embodiment, actual users in different groups can be migrated between groups based on the influence factor score of each actual user for each influence factor. For example, for each influence factor, actual users with small influence factor scores are migrated to groups with larger influence factor scores to improve overall satisfaction, and the migration score of each influence factor after migration is determined. The migration score represents the degree to which the influence factor improves the overall satisfaction. The larger the migration score, the greater the improvement in overall satisfaction.

[0097] In step S105, the impact factor corresponding to the maximum migration score is determined as the target impact factor.

[0098] In this embodiment, the impact factor corresponding to the maximum migration score is determined as the target impact factor, so that more effective business adjustment measures can be formulated for the target impact factor, thereby maximizing the overall user satisfaction.

[0099] In this embodiment, by obtaining multiple actual data corresponding to multiple actual users, and determining the target grouping of each actual user with respect to each influencing factor based on the multiple actual data, and then training the basic network with multiple sample overall satisfaction and multiple sample influencing factor satisfaction corresponding to multiple sample users to obtain an overall satisfaction prediction model, and constructing a score mapping model based on the overall satisfaction prediction model, the target grouping of each actual user with respect to each influencing factor is processed according to the score mapping model, and the influencing factor score of each actual user with respect to each influencing factor is obtained, and the migration score of each influencing factor is further determined, thereby determining the target influencing factor with the largest migration score. Through the above method, the quantitative relationship between each influencing factor and the overall satisfaction can be reflected through the influencing factor score of each actual user with respect to each influencing factor, thereby determining the target influencing factor that can maximize the improvement of the overall satisfaction, so as to implement business adjustment measures according to the target influencing factor, thereby maximizing the overall satisfaction of the user.

[0100] Figure 2 is a flowchart of a method for training an overall satisfaction prediction model according to an exemplary embodiment. Figure 2 As shown, in a possible implementation, the overall satisfaction prediction model can be trained by the following steps:

[0101] In step S201, a plurality of sample data corresponding to a plurality of sample users are obtained, the sample data including the sample overall satisfaction and the sample influencing factor satisfaction of a plurality of influencing factors associated with the sample overall satisfaction.

[0102] In this embodiment, multiple sample users may be users related to the target service, and the multiple sample data may be obtained through a questionnaire survey. For example, a satisfaction questionnaire may be sent to multiple sample users. The satisfaction questionnaire may be a total score structure. The satisfaction questionnaire may inquire about the overall satisfaction and the satisfaction of multiple influencing factors associated with the overall satisfaction, such as logistics progress inquiry satisfaction and logistics timeliness satisfaction, etc. The overall satisfaction and the satisfaction of the influencing factors are both interval values, such as integer values ​​between 1 and 10. Influencing factors can affect the overall satisfaction of users. For example, the influencing factors may be shipping warehouses, carriers, shipping locations, receiving locations, purchase channels, outbound time, and delivery time, etc. The overall satisfaction may be associated with multiple influencing factors. Through the satisfaction questionnaire, the overall sample satisfaction and multiple sample influencing factor satisfaction corresponding to each sample user may be obtained.

[0103] In step S202, the overall satisfaction label of each sample user is determined according to a preset overall satisfaction threshold and the overall satisfaction of the sample.

[0104] In this embodiment, the preset overall satisfaction threshold can be a pre-set value used to distinguish between satisfaction and dissatisfaction. For example, the preset overall satisfaction threshold is 9, and the sample with an overall satisfaction less than 9 is dissatisfied, and greater than or equal to 9 is satisfied. A sample user has a satisfaction label corresponding to the overall satisfaction, and satisfaction is determined as 1, otherwise it is 0. The satisfaction label is 0 or 1. If the overall satisfaction is greater than or equal to 9, it is determined as 1, otherwise it is 0.

[0105] In step S203, the influence factor satisfaction of multiple influence factors of multiple sample users is discretized to obtain an original sample grouping of each sample user with respect to each influence factor.

[0106] In this embodiment, the influence factor satisfaction of each influencing factor can be discretized. Optionally, the grouping basis can be based on equal frequency grouping, and the equal frequency grouping is that the number of samples in the divided interval is basically the same, so as to determine the original grouping parameters. The grouping parameter can be at least one grouping threshold, and the sample business parameters of each influencing factor can be divided into multiple groups by at least one grouping threshold. Grouping can also be performed according to the size of the influence factor satisfaction. For example, the grouping threshold can be 2, 6 and 9, and the actual influence factor satisfaction is between 0 and 2 as a group, between 2 and 6 as a group, between 6 and 9 as a group, and between 9 and 10 as a group. Through the original grouping parameters, the original sample grouping of each sample user for each influencing factor can be obtained.

[0107] In step S204, the evidence weight value of each original sample group is obtained according to the overall satisfaction label of each sample user.

[0108] In this embodiment, the calculation formula of the weight of evidence value can be:

[0109]

[0110] Among them, woe ij is the weight of evidence for the jth group of the ith influencing factor, p ij1 is the proportion of satisfied users in the jth group of the ith influencing factor to all satisfied users, p ij0 is the proportion of dissatisfied users in the jth group of the ith influencing factor to all satisfied users, S ij is the number of satisfied users in the jth group of the ith influencing factor, ST is the total number of satisfied users, UNS is ij is the number of dissatisfied users in the jth group of the i-th influencing factor, and UNST is the total number of dissatisfied customers. Among them, satisfied and dissatisfied are both satisfied and dissatisfied with the overall satisfaction. The above parameters can be obtained based on the satisfaction label of each sample user for the overall satisfaction.

[0111] In step S205, the basic network is trained for multiple rounds according to the evidence weight values ​​of the original sample groupings of each sample user with respect to multiple influencing factors and the overall satisfaction label of each sample user to obtain an overall satisfaction prediction model.

[0112] In this embodiment, the basic network may be a Logistic regression model, and the prediction formula of the model may be:

[0113]

[0114] Among them, α is the intercept term of the overall satisfaction prediction model, β i is the slope of the i-th influencing factor, woe ij is the weight of evidence for the jth group of the ith influencing factor, σ ij Whether to take the jth group of the i-th influencing factor is a binary variable, which is 1 if yes and 0 otherwise. For a factor i, only j can be taken at a time.

[0115] According to the evidence weight value of each sample user's original sample grouping of the multiple influencing factors and the overall satisfaction label of each sample user, the basic network is trained for multiple rounds. After each round of training, the predicted probability value output by the basic network is obtained, and the predicted probability value ranges from 0 to 1. And according to the overall satisfaction label and the predicted probability value, the loss value is determined, and the basic network is optimized based on the loss value. Until the basic network meets the preset condition, the preset condition can be that the basic network converges or the number of training rounds reaches the preset number of rounds, the training can be stopped to obtain the overall satisfaction prediction model.

[0116] In a possible implementation, before performing multiple rounds of training on the basic network based on the evidence weight values ​​of the original sample groupings of each sample user regarding multiple influencing factors and the overall satisfaction label of each sample user to obtain an overall satisfaction prediction model, the target influencing factor determination method further includes: screening multiple influencing factors based on the evidence weight values ​​of each original sample grouping to obtain initial influencing factors and multiple updated sample groupings corresponding to each initial influencing factor; and obtaining the evidence weight value of each updated sample group based on the overall satisfaction label of each sample user.

[0117] In this embodiment, before training the model, multiple influencing factors can be screened in advance. For example, influencing factors that do not have monotonicity can be deleted, and influencing factors whose monotonicity is opposite to the expected monotonicity can be deleted. Here, the expected monotonicity is monotonically increasing, so as to improve the accuracy of the overall satisfaction prediction model obtained by training. Optionally, whether each influencing factor is monotonic and the specific monotonic direction can be determined based on the evidence weight value of each original sample grouping. The monotonic direction can be monotonically increasing or monotonically decreasing. Then, multiple influencing factors are screened to obtain initial influencing factors and multiple updated sample groups corresponding to each initial influencing factor. Since the multiple updated sample groups corresponding to each initial influencing factor may have group changes relative to the original sample groups, the evidence weight value of each updated sample grouping can be re-determined based on the overall satisfaction label of each sample user.

[0118] After obtaining the initial influencing factors and the evidence weight values ​​of the updated sample groups, the basic network can be trained multiple times according to the evidence weight values ​​of the updated sample groups of each sample user regarding the initial influencing factors and the overall satisfaction labels of each sample user to obtain an overall satisfaction prediction model.

[0119] In this embodiment, by pre-screening multiple influencing factors, the accuracy of the trained overall satisfaction prediction model can be improved.

[0120] Figure 3is a flow chart of a method for screening impact factors according to an exemplary embodiment. Figure 3 As shown, in a possible implementation, according to the evidence weight value of each original sample group, multiple influencing factors are screened to obtain initial influencing factors and multiple updated sample groups corresponding to each initial influencing factor, which may include the following steps:

[0121] In step S301, for any influencing factor among the multiple influencing factors, according to the multiple original sample groups corresponding to the any influencing factor, an actual monotonic relationship between the evidence weight values ​​of the multiple original sample groups is determined.

[0122] In this embodiment, the actual monotonic relationship of the evidence weight values ​​of the multiple original sample groups corresponding to any influencing factor can be determined according to the evidence weight values ​​of the multiple original sample groups, and the actual monotonic relationship of any influencing factor can be obtained. The actual monotonic relationship can be divided into non-monotonic, monotonic increasing and monotonic decreasing.

[0123] In step S302, when the actual monotonic relationship is monotonically increasing, any influencing factor is determined as an initial influencing factor, and the sample grouping in the monotonically increasing state is determined as a plurality of updated sample groups corresponding to the initial influencing factor.

[0124] In this embodiment, the expected monotonic relationship is monotonically increasing to ensure that the influencing factor is positively correlated with the overall satisfaction. When the actual monotonic relationship is monotonically increasing, any influencing factor is determined as an initial influencing factor, and the sample grouping during monotonically increasing period is determined as multiple updated sample groups corresponding to the initial influencing factor.

[0125] In this embodiment, the influencing factors with a monotonic relationship and monotonically increasing can be determined as initial influencing factors, so that the data corresponding to the initial influencing factors are used for model training to improve the accuracy of the overall satisfaction prediction model obtained through training.

[0126] Figure 4 is a flow chart showing a method for determining an actual monotonic relationship according to an exemplary embodiment. Figure 4 As shown, in a possible implementation, for any influencing factor among the multiple influencing factors, according to the multiple original sample groups corresponding to the any influencing factor, determining the actual monotonic relationship of the evidence weight values ​​of the multiple original sample groups may include the following steps:

[0127] In step S401, a plurality of original sample groups corresponding to any influencing factor are determined as a plurality of basic groups.

[0128] In step S402, multiple basic groups are sorted, and the sequence number of each basic group is determined.

[0129] In this implementation, optionally, multiple basic groups may be sorted in ascending order according to the size of the sample service parameters in the basic groups, and the sorted sequence number may be determined as the sequence number of each basic group.

[0130] In step S403, when the monotonicity of the evidence weight values ​​of the sorted plurality of basic groups is non-monotonic, the target sequence number of the basic group corresponding to the non-monotonic target evidence weight value is determined.

[0131] In this embodiment, when there are multiple basic groups corresponding to non-monotonic target evidence weight values, the sequence number of one basic group can be arbitrarily selected from the multiple non-monotonic basic groups as the target sequence number.

[0132] In step S404, multiple basic groups are merged according to the target sequence number to obtain a first group sequence.

[0133] In this implementation, based on the target sequence number, basic groups with adjacent sequence numbers may be merged to obtain a merged first group sequence.

[0134] Optionally, when the target sequence number is the first sequence number, based on the group corresponding to the target sequence number, basic groups of adjacent sequence numbers are merged backward to obtain a first group sequence. Merging basic groups of adjacent sequence numbers backward is to merge basic groups of adjacent sequence numbers that are sorted later.

[0135] Optionally, when the target sequence number is the last sequence number, based on the group corresponding to the target sequence number, basic groups of adjacent sequence numbers are merged forward to obtain a first group sequence. Merging basic groups of adjacent sequence numbers forward is to merge basic groups of adjacent sequence numbers that are sorted forward.

[0136] Optionally, the first group sequence may include a first group subsequence and a second group subsequence. When the target sequence number is an intermediate sequence number, the basic groups of adjacent sequence numbers are merged backward based on the group corresponding to the target sequence number to obtain the first group subsequence, and the basic groups of adjacent sequence numbers are merged forward based on the group corresponding to the target sequence number to obtain the second group subsequence.

[0137] In step S405, when the monotonicity of the evidence weight values ​​of the first grouping sequence is non-monotonic, multiple new basic groups are determined based on the first grouping sequence, and the evidence weight value of each new basic group is determined until the number of the multiple new basic groups is less than the preset number of groups, or the evidence weight values ​​of the multiple new basic groups are monotonic.

[0138] In this embodiment, if the monotonicity of the evidence weight values ​​of the merged first grouping sequence is still non-monotonic, the first grouping sequence may be determined as a plurality of new basic groups, and steps S402 to S405 may be repeatedly performed on the new basic groups until the number of the plurality of new basic groups is less than the preset number of groups, or the evidence weight values ​​of the plurality of basic groups are monotonic, wherein the preset number of groups may be 2. Optionally, if the number of the plurality of new basic groups is less than the preset number of groups, it is determined that the actual monotonic relationship of the evidence weight values ​​of the plurality of original sample groups is non-monotonic, and the corresponding influencing factors may be deleted and not used as initial influencing factors.

[0139] In step S406, the basic grouping when the evidence weight value is monotonic is determined as a plurality of updated sample groups.

[0140] In this embodiment, the basic grouping when the evidence weight value is monotonic can be: a first grouping sequence whose evidence weight value is monotonic, or a plurality of new basic groups whose evidence weight value is monotonic. The plurality of groups included in the first grouping sequence whose evidence weight value is monotonic are determined as a plurality of updated sample groups, or the plurality of new basic groups whose evidence weight value is monotonic are determined as a plurality of updated sample groups.

[0141] In step S407, the monotonic relationship between the evidence weight values ​​of the multiple updated sample groups is determined as the actual monotonic relationship between the evidence weight values ​​of the multiple original sample groups.

[0142] In this embodiment, after obtaining multiple updated sample groups, the monotonic relationship of the evidence weight values ​​of the multiple updated sample groups can be determined, and the monotonic relationship is determined as the actual monotonic relationship of the evidence weight values ​​of the multiple original sample groups. The actual monotonic relationship can be monotonically increasing or monotonically decreasing.

[0143] In this embodiment, multiple non-monotonic groups are merged through the monotonic combination box method until the evidence weight values ​​of the multiple groups are monotonic, so that more reasonable training samples can be obtained, so that the overall satisfaction prediction model obtained by training is more accurate.

[0144] In one possible implementation, when the monotonicity of the evidence weight values ​​of the first grouping sequence is non-monotonic, the method for determining multiple new basic groupings based on the first grouping sequence may be: when the monotonicity of the evidence weight values ​​of the first grouping subsequence and the second grouping subsequence are both non-monotonic, determine the first information value of the influencing factor corresponding to the first grouping subsequence, and determine the second information value of the influencing factor corresponding to the second grouping subsequence; determine the grouping subsequences corresponding to the larger value of the first information value and the second information value as multiple new basic groupings.

[0145] In this embodiment, when the first grouping sequence includes the first grouping subsequence and the second grouping subsequence, the grouping subsequences with greater information value are determined as multiple new basic groups. When there is only one grouping sequence in the first grouping sequence, the first grouping sequence is directly determined as a multiple basic grouping.

[0146] Among them, the calculation formula for the information value of the jth group of the i-th influencing factor is:

[0147]

[0148] The calculation formula for the information value of the i-th impact factor is:

[0149]

[0150] The meaning of the characters in the above formula can be referred to the above content, which will not be repeated here. The range of information value is between 0 and 1.

[0151] Figure 5 is a flow chart showing a method for determining a second target impact factor according to an exemplary embodiment. Figure 5 As shown, in a possible implementation, before the basic network is trained multiple times according to the evidence weight value of the updated sample grouping of each sample user with respect to the initial influencing factor and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model, the target influencing factor determination method further includes:

[0152] In step S501, the third information value of each initial impact factor is determined.

[0153] In this implementation, the third information value of each initial impact factor can be determined by the above-mentioned information value calculation method.

[0154] In step S502, an initial impact factor whose third information value is greater than a preset information value threshold is determined as a first target impact factor.

[0155] In this embodiment, optionally, the preset information value threshold may be 0.25, that is, the initial impact factor with a third information value greater than 0.25 may be determined as the first target impact factor.

[0156] In step S503, the correlation coefficient between any two first target impact factors is determined according to the evidence weight value of each updated sample group corresponding to each first target impact factor.

[0157] In this embodiment, the correlation coefficient may be a Pearson correlation coefficient, and its calculation formula may be:

[0158]

[0159] Where p is the number of sample users, p∈[1,n]; r,q are the impact factors, r≠q, r∈[1,m], q∈[1,m]; woe rp is the woe value of the influencing factor g in sample p, i.e., the weight of evidence value; is the mean value of woe of the influencing factor r; woe qp is the impact factor q, the woe value in sample p; is the woe mean of the influencing factor q.

[0160] In step S504, a screening influencing factor is determined according to the correlation coefficient, where the screening influencing factor is the first target influencing factor with the smaller third information value among the two first target influencing factors whose correlation coefficient is greater than a preset correlation coefficient threshold.

[0161] In this embodiment, if the correlation coefficient between any two first target influencing factors is greater than a preset correlation coefficient threshold, for example, the preset correlation coefficient threshold may be 0.8, then the two first target influencing factors are similar, and only one may be selected for model training. Optionally, the first target influencing factor with the smaller third information value among the two first target influencing factors may be determined as the screening influencing factor.

[0162] In step S505, the screening impact factor is deleted from the first target impact factor to obtain a second target impact factor.

[0163] In this embodiment, by deleting the screening impact factor from the first target impact factor, a more reasonable second target impact factor can be obtained.

[0164] After obtaining the second target influencing factor, the basic network can be trained for multiple rounds according to the evidence weight value of the updated sample grouping of each sample user regarding the second target influencing factor and the overall satisfaction label of each sample user to obtain an overall satisfaction prediction model.

[0165] In this embodiment, the influencing factors with smaller information value in the initial influencing factors can be deleted, and the influencing factors with smaller information value in the two more relevant influencing factors can be deleted to obtain the second target influencing factors, so that a more accurate overall satisfaction prediction model can be obtained through training with the second target influencing factors.

[0166] In a possible implementation, the method for constructing the score mapping model may be: obtaining an overall satisfaction prediction model, and constructing a score mapping model based on the overall satisfaction prediction model, according to a target mapping interval and a target mapping rule.

[0167] The target mapping rule may be a linear mapping relationship between overall satisfaction and overall satisfaction score, and the formula for the overall satisfaction score is:

[0168]

[0169] And obtain the target mapping interval, which is the numerical interval of the overall satisfaction score, which can be a pre-set interval. The target mapping interval is determined by the maximum score and the minimum score. For example, the minimum score min_score can be 0, the maximum score max_score can be 10, and the target mapping interval can be [0,10].

[0170] Then, the offset and factor can be obtained through the formula of the overall satisfaction score and the target mapping interval.

[0171] in,

[0172]

[0173] And based on the calculation formula of overall satisfaction:

[0174]

[0175] The calculation formula of the score mapping model is:

[0176]

[0177] The score mapping model can determine the corresponding values ​​of i and j according to the target grouping of each actual user for each influencing factor, and then determine the value of each parameter in the above formula, so as to obtain the influencing factor score of each actual user for each influencing factor.

[0178] Among them, the starting score of customer satisfaction is: offset+factor×α, and the single score of influencing factor i in group j is: factor*βi*woeij*σij

[0179] The score sample table of the score mapping model is as follows:

[0180]

[0181] Figure 6 is a flow chart showing a method for determining a migration score according to an exemplary embodiment. Figure 6 As shown, in a possible implementation, determining the migration score of each influencing factor according to the influencing factor score of each actual user with respect to each influencing factor may include the following steps:

[0182] In step S601 , based on the influence factor score of each actual user with respect to each influence factor, the influence factor score corresponding to each group of each influence factor and the number of actual users corresponding to each group are determined.

[0183] In this embodiment, according to the influence factor score of each actual user for each influence factor, it is possible to determine which target group the actual user belongs to for the influence factor and the influence factor score corresponding to the target group, thereby determining the influence factor score corresponding to each group of each influence factor and the number of actual users corresponding to each group. For example, for influence factor 1, its grouping includes group 1, group 2, and group 3, and the influence factor scores corresponding to each group are: factor*β1*woe11, factor*β1*woe12, and factor*β1*woe13, respectively, and the number of actual users corresponding to each group is p11, p12, and p13.

[0184] In step S602, the migration group and the target group of each impact factor are determined according to the impact factor score corresponding to each group of each impact factor, and the impact factor score corresponding to the target group is greater than the impact factor score corresponding to the migration group.

[0185] In this embodiment, users can be migrated between groups according to the impact factor scores corresponding to each group of each impact factor, so as to improve the overall satisfaction. Optionally, the group with the largest score can be determined as the target group, and the group with the smallest score can be determined as the migration group. For example, group 1 can be determined as the migration group, and group 3 can be determined as the target group.

[0186] In step S603, the migration score of each influencing factor is determined according to the influencing factor score corresponding to the migration group of each influencing factor, the influencing factor score corresponding to the target group, and the number of actual users corresponding to the migration group.

[0187] In this embodiment, for any impact factor, the difference between the impact factor score corresponding to the target group of the impact factor and the impact factor score corresponding to the migration group can be multiplied by the number of actual users corresponding to the migration group to obtain the migration score. For example, for impact factor 1, its migration score can be: (factor*β 1 *woe 13 -factor*β 1 *woe 11 )*p 11 point.

[0188] Figure 7 is a block diagram of a device for determining a target impact factor according to an exemplary embodiment. Figure 7The device includes a first acquisition module 701, a discrete module 702, a first acquisition module 703, a first determination module 704 and a second determination module 705.

[0189] The first acquisition module 701 is configured to acquire a plurality of actual data corresponding to a plurality of actual users, wherein the actual data includes an actual overall satisfaction degree and actual influencing factor satisfaction degrees of a plurality of influencing factors associated with the actual overall satisfaction degree;

[0190] The discretization module 702 is configured to discretize the actual influencing factor satisfaction of the multiple influencing factors of the multiple actual users to obtain a target grouping of each actual user with respect to each influencing factor;

[0191] The first obtaining module 703 is configured to process the target grouping of each actual user with respect to each influencing factor according to the score mapping model, respectively, to obtain the influencing factor score of each actual user with respect to each influencing factor, wherein the score mapping model is constructed based on the overall satisfaction prediction model, and the overall satisfaction prediction model is obtained by training a basic network based on multiple sample overall satisfactions and multiple sample influencing factor satisfactions corresponding to multiple sample users;

[0192] The first determination module 704 is configured to determine a migration score of each influencing factor according to the influencing factor score of each actual user with respect to each influencing factor, wherein the migration score represents the degree to which the influencing factor improves the overall satisfaction;

[0193] The second determination module 705 is configured to determine the impact factor corresponding to the maximum migration score as the target impact factor.

[0194] Optionally, the device further comprises:

[0195] A second acquisition module is configured to acquire a plurality of sample data corresponding to a plurality of sample users, wherein the sample data includes a sample overall satisfaction and a sample influencing factor satisfaction of a plurality of influencing factors associated with the sample overall satisfaction;

[0196] A third determination module is configured to determine an overall satisfaction label of each sample user according to a preset overall satisfaction threshold and the sample overall satisfaction;

[0197] A second obtaining module is configured to discretize the influence factor satisfaction of the multiple influence factors of the multiple sample users to obtain an original sample grouping of each sample user with respect to each influence factor;

[0198] The third obtaining module is configured to obtain the evidence weight value of each original sample grouping according to the overall satisfaction label of each sample user;

[0199] The fourth acquisition module is configured to perform multiple rounds of training on the basic network according to the evidence weight value of each sample user's original sample grouping regarding the multiple influencing factors and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model.

[0200] Optionally, the device further comprises:

[0201] A screening module is configured to screen the multiple influencing factors according to the evidence weight value of each original sample group to obtain an initial influencing factor and multiple updated sample groups corresponding to each initial influencing factor;

[0202] a fifth obtaining module, configured to obtain an evidence weight value of each updated sample grouping according to the overall satisfaction label of each sample user;

[0203] The fourth acquisition module comprises:

[0204] The acquisition submodule is configured to perform multiple rounds of training on the basic network according to the evidence weight value of the updated sample grouping of each sample user with respect to the initial influencing factor and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model.

[0205] Optionally, the screening module includes:

[0206] A first determination submodule is configured to determine, for any influencing factor among the multiple influencing factors, an actual monotonic relationship of the evidence weight values ​​of the multiple original sample groups according to the multiple original sample groups corresponding to the any influencing factor;

[0207] The second determination submodule is configured to determine any one of the influencing factors as an initial influencing factor when the actual monotonic relationship is monotonically increasing, and determine the monotonically increasing sample grouping as multiple updated sample groups corresponding to the initial influencing factor.

[0208] Optionally, the second determining submodule includes:

[0209] A first determining unit is configured to determine a plurality of original sample groups corresponding to any one of the influencing factors as a plurality of basic groups;

[0210] A sorting unit, configured to sort the multiple basic groups and determine a sequence number of each basic group;

[0211] A second determining unit is configured to determine a target sequence number of a basic group corresponding to a non-monotonic target evidence weight value when the monotonicity of the evidence weight values ​​of the sorted plurality of basic groups is non-monotonic;

[0212] a merging unit, configured to merge the multiple basic groups according to the target sequence number to obtain a first group sequence;

[0213] a third determining unit configured to determine a plurality of new basic groups according to the first grouping sequence, and determine the evidence weight value of each new basic grouping, if the monotonicity of the evidence weight value of the first grouping sequence is not monotonic, until the number of the plurality of new basic groups is less than the preset number of groups, or the evidence weight values ​​of the plurality of new basic groups are monotonic;

[0214] A fourth determining unit is configured to determine the basic grouping when the evidence weight value is monotonic as a plurality of update sample groups;

[0215] A fifth determining unit is configured to determine the monotonic relationship among the evidence weight values ​​of the multiple updated sample groups as the actual monotonic relationship among the evidence weight values ​​of the multiple original sample groups.

[0216] Optionally, the merging unit includes:

[0217] The first merging subunit is configured to, when the target sequence number is the first sequence number, backward merge the basic groups of adjacent sequence numbers based on the group corresponding to the target sequence number to obtain the first group sequence.

[0218] Optionally, the merging unit includes:

[0219] The second merging subunit is configured to, when the target sequence number is the last sequence number, forward merge the basic groups of adjacent sequence numbers based on the group corresponding to the target sequence number to obtain the first group sequence.

[0220] Optionally, the first packet sequence includes a first packet subsequence and a second packet subsequence;

[0221] The merging unit comprises:

[0222] The third merging subunit is configured to, when the target sequence number is an intermediate sequence number, merge the basic groups of adjacent sequence numbers backward based on the group corresponding to the target sequence number to obtain a first group subsequence, and merge the basic groups of adjacent sequence numbers forward based on the group corresponding to the target sequence number to obtain a second group subsequence.

[0223] Optionally, the third determining unit includes:

[0224] A first determining subunit is configured to determine a first information value of an influencing factor corresponding to the first grouping subsequence and a second information value of an influencing factor corresponding to the second grouping subsequence when the monotonicity of the evidence weight values ​​of the first grouping subsequence and the second grouping subsequence are both non-monotonic;

[0225] The second determining subunit is configured to determine the grouping subsequence corresponding to the larger value of the first information value and the second information value as a plurality of new basic groups.

[0226] Optionally, the device further comprises:

[0227] a fourth determination module configured to determine a third information value of each initial impact factor;

[0228] A fifth determination module is configured to determine the initial impact factor whose third information value is greater than a preset information value threshold as a first target impact factor;

[0229] A sixth determination module is configured to determine a correlation coefficient between any two first target impact factors according to the evidence weight value of each updated sample group corresponding to each first target impact factor;

[0230] A seventh determination module is configured to determine a screening influence factor according to the correlation coefficient, wherein the screening influence factor is the first target influence factor with a smaller third information value among the two first target influence factors whose correlation coefficient is greater than a preset correlation coefficient threshold;

[0231] a sixth obtaining module, configured to delete the screening impact factor from the first target impact factor to obtain a second target impact factor;

[0232] The obtaining submodule comprises:

[0233] The obtaining unit is configured to perform multiple rounds of training on the basic network according to the evidence weight value of the updated sample grouping of each sample user regarding the second target influencing factor and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model.

[0234] Optionally, the first determining module includes:

[0235] A third determination submodule is configured to determine the influence factor score corresponding to each group of each influence factor and the number of actual users corresponding to each group according to the influence factor score of each actual user with respect to each influence factor;

[0236] A fourth determination submodule is configured to determine a migration group and a target group of each impact factor according to the impact factor score corresponding to each group of each impact factor, wherein the impact factor score corresponding to the target group is greater than the impact factor score corresponding to the migration group;

[0237] The fifth determination submodule is configured to determine the migration score of each influencing factor according to the influencing factor score corresponding to the migration group of each influencing factor, the influencing factor score corresponding to the target group, and the number of actual users corresponding to the migration group.

[0238] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0239] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the steps of the target impact factor determination method provided by the present disclosure.

[0240] Figure 8 FIG. 8 is a block diagram of another target impact factor determination device according to an exemplary embodiment. For example, device 800 may be provided as a server. Figure 8 The apparatus 800 includes a processing component 822, which further includes one or more processors, and a memory resource represented by a memory 832 for storing instructions executable by the processing component 822, such as an application. The application stored in the memory 832 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 822 is configured to execute instructions to perform the above-mentioned target impact factor determination method.

[0241] The device 800 may also include a power supply component 826 configured to perform power management of the device 800, a wired or wireless network interface 850 configured to connect the device 800 to a network, and an input / output interface 858. The device 800 may operate based on an operating system stored in the memory 832, such as Windows Server 2000. TM , Mac OS X TM , Unix TM , Linux TM , FreeBSD TM or similar.

[0242] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the present disclosure. The present disclosure is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0243] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for determining a target impact factor, characterized in that: include: Acquire a plurality of actual data corresponding to a plurality of actual users, the actual data including actual overall satisfaction and actual influencing factor satisfaction of a plurality of influencing factors associated with the actual overall satisfaction; Discretize the actual influencing factor satisfaction of the multiple influencing factors of the multiple actual users to obtain a target grouping of each actual user with respect to each influencing factor; The target grouping of each actual user with respect to each influencing factor is processed respectively according to the score mapping model to obtain the influencing factor score of each actual user with respect to each influencing factor, wherein the score mapping model is constructed based on the overall satisfaction prediction model, and the overall satisfaction prediction model is obtained by training a basic network based on multiple sample overall satisfactions and multiple sample influencing factor satisfactions corresponding to multiple sample users; Determine a migration score for each influencing factor based on the influencing factor score of each actual user for each influencing factor, wherein the migration score represents the degree to which the influencing factor improves overall satisfaction; The impact factor corresponding to the maximum migration score is determined as the target impact factor.

2. The method for determining the target impact factor according to claim 1, characterized in that: The overall satisfaction prediction model is trained by the following steps: Acquire a plurality of sample data corresponding to a plurality of sample users, wherein the sample data includes sample overall satisfaction and sample influencing factor satisfaction of a plurality of influencing factors associated with the sample overall satisfaction; Determine an overall satisfaction label for each sample user based on a preset overall satisfaction threshold and the overall satisfaction of the sample; Discretizing the influence factor satisfaction of the multiple influence factors of the multiple sample users to obtain an original sample grouping of each sample user with respect to each influence factor; According to the overall satisfaction label of each sample user, the evidence weight value of each original sample group is obtained; According to the evidence weight value of each sample user's original sample grouping with respect to the multiple influencing factors and the overall satisfaction label of each sample user, the basic network is trained for multiple rounds to obtain the overall satisfaction prediction model.

3. The method for determining the target impact factor according to claim 2, characterized in that: Before performing multiple rounds of training on the basic network according to the evidence weight values ​​of the original sample groupings of each sample user with respect to the multiple influencing factors and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model, the method further includes: According to the evidence weight value of each original sample group, the multiple influencing factors are screened to obtain initial influencing factors and multiple updated sample groups corresponding to each initial influencing factor; According to the overall satisfaction label of each sample user, the evidence weight value of each updated sample group is obtained; The method of performing multiple rounds of training on the basic network according to the evidence weight value of each sample user's original sample grouping with respect to the multiple influencing factors and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model includes: According to the evidence weight value of each sample user's updated sample grouping with respect to the initial influencing factor and the overall satisfaction label of each sample user, the basic network is trained for multiple rounds to obtain the overall satisfaction prediction model.

4. The method for determining the target impact factor according to claim 3, characterized in that: The method of screening the multiple influencing factors according to the evidence weight value of each original sample group to obtain the initial influencing factors and multiple updated sample groups corresponding to each initial influencing factor includes: For any influencing factor among the multiple influencing factors, according to the multiple original sample groups corresponding to the any influencing factor, determine the actual monotonic relationship of the evidence weight values ​​of the multiple original sample groups; In the case where the actual monotonic relationship is monotonically increasing, any one of the influencing factors is determined to be an initial influencing factor, and the sample groupings in the monotonically increasing state are determined to be a plurality of updated sample groups corresponding to the initial influencing factors.

5. The method for determining the target impact factor according to claim 4, characterized in that: The determining, for any influencing factor among the multiple influencing factors, according to the multiple original sample groups corresponding to the any influencing factor, an actual monotonic relationship of the evidence weight values ​​of the multiple original sample groups comprises: Determine a plurality of original sample groups corresponding to any of the influencing factors as a plurality of basic groups; Sorting the multiple basic groups and determining a sequence number for each basic group; When the monotonicity of the evidence weight values ​​of the plurality of basic groups after sorting is non-monotonic, determining a target sequence number of the basic group corresponding to the non-monotonic target evidence weight value; According to the target sequence number, the multiple basic groups are combined to obtain a first group sequence; In the case where the monotonicity of the evidence weight values ​​of the first grouping sequence is not monotonic, a plurality of new basic groups are determined according to the first grouping sequence, and the evidence weight value of each new basic group is determined until the number of the plurality of new basic groups is less than the preset number of groups, or the evidence weight values ​​of the plurality of new basic groups are monotonic; The basic grouping when the evidence weight value is monotonic is determined as a plurality of updated sample groups; The monotonic relationship among the evidence weight values ​​of the multiple updated sample groups is determined as the actual monotonic relationship among the evidence weight values ​​of the multiple original sample groups.

6. The method for determining the target impact factor according to claim 5, characterized in that: The step of merging the plurality of basic groups according to the target sequence number to obtain a first group sequence includes: In the case where the target sequence number is the first sequence number, based on the group corresponding to the target sequence number, basic groups of adjacent sequence numbers are merged backward to obtain the first group sequence.

7. The method for determining the target impact factor according to claim 5, characterized in that: The step of merging the plurality of basic groups according to the target sequence number to obtain a first group sequence includes: In the case where the target sequence number is the last sequence number, based on the group corresponding to the target sequence number, basic groups of adjacent sequence numbers are merged forward to obtain the first group sequence.

8. The method for determining the target impact factor according to claim 5, characterized in that: The first packet sequence includes a first packet subsequence and a second packet subsequence; The step of merging the plurality of basic groups according to the target sequence number to obtain a first group sequence includes: When the target sequence number is an intermediate sequence number, the basic groups of adjacent sequence numbers are merged backward based on the group corresponding to the target sequence number to obtain a first group subsequence, and the basic groups of adjacent sequence numbers are merged forward based on the group corresponding to the target sequence number to obtain a second group subsequence.

9. The method for determining the target impact factor according to claim 8, characterized in that: The step of determining a plurality of new basic groups according to the first grouping sequence when the monotonicity of the evidence weight values ​​of the first grouping sequence is non-monotonic comprises: In a case where the monotonicity of the evidence weight values ​​of the first grouping subsequence and the second grouping subsequence are both non-monotonic, determining a first information value of an influencing factor corresponding to the first grouping subsequence, and determining a second information value of an influencing factor corresponding to the second grouping subsequence; The group subsequences corresponding to the larger value of the first information value and the second information value are determined as a plurality of new basic groups.

10. The method for determining a target impact factor according to any one of claims 3 to 9, characterized in that: Before performing multiple rounds of training on the basic network according to the evidence weight value of the updated sample grouping of each sample user with respect to the initial influencing factor and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model, the method further includes: Determine the third information value of each initial impact factor; Determine the initial impact factor whose third information value is greater than the preset information value threshold as the first target impact factor; Determine the correlation coefficient between any two first target impact factors according to the evidence weight value of each updated sample group corresponding to each first target impact factor; Determine a screening influence factor according to the correlation coefficient, wherein the screening influence factor is the first target influence factor with the smaller third information value among the two first target influence factors whose correlation coefficient is greater than a preset correlation coefficient threshold; Deleting the screening impact factor from the first target impact factor to obtain a second target impact factor; The method of performing multiple rounds of training on the basic network according to the evidence weight value of the updated sample grouping of each sample user with respect to the initial influencing factor and the overall satisfaction label of each sample user to obtain the overall satisfaction prediction model includes: According to the evidence weight value of the updated sample grouping of each sample user regarding the second target influencing factor and the overall satisfaction label of each sample user, the basic network is trained for multiple rounds to obtain the overall satisfaction prediction model.

11. The method for determining a target impact factor according to any one of claims 1 to 9, characterized in that: Determining the migration score of each influencing factor according to the influencing factor score of each actual user with respect to each influencing factor includes: Determine, according to the influence factor score of each actual user with respect to each influence factor, the influence factor score corresponding to each group of each influence factor and the number of actual users corresponding to each group; Determine a migration group and a target group for each impact factor according to the impact factor score corresponding to each group of each impact factor, wherein the impact factor score corresponding to the target group is greater than the impact factor score corresponding to the migration group; The migration score of each influencing factor is determined according to the influencing factor score corresponding to the migration group of each influencing factor, the influencing factor score corresponding to the target group, and the number of actual users corresponding to the migration group.

12. A device for determining a target impact factor, characterized in that: include: A first acquisition module is configured to acquire a plurality of actual data corresponding to a plurality of actual users, wherein the actual data includes an actual overall satisfaction degree and actual influencing factor satisfaction degrees of a plurality of influencing factors associated with the actual overall satisfaction degree; A discretization module is configured to discretize the actual influencing factor satisfaction of the multiple influencing factors of the multiple actual users to obtain a target grouping of each actual user with respect to each influencing factor; A first obtaining module is configured to process the target grouping of each actual user with respect to each influencing factor respectively according to a score mapping model to obtain an influencing factor score of each actual user with respect to each influencing factor, wherein the score mapping model is constructed based on an overall satisfaction prediction model, and the overall satisfaction prediction model is obtained by training a basic network based on a plurality of sample overall satisfactions and a plurality of sample influencing factor satisfactions corresponding to a plurality of sample users; A first determination module is configured to determine a migration score of each influencing factor according to the influencing factor score of each actual user with respect to each influencing factor, wherein the migration score represents the degree to which the influencing factor improves the overall satisfaction; The second determination module is configured to determine the impact factor corresponding to the maximum migration score as the target impact factor.

13. A device for determining a target impact factor, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to implement the steps of the target impact factor determination method according to any one of claims 1 to 11.

14. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by a processor, the steps of the target impact factor determination method described in any one of claims 1 to 11 are implemented.

Images