User attribution prediction method, device and equipment based on model adaptive adjustment

Abstract

The present invention discloses a user attribution prediction method, device and equipment based on model adaptive adjustment, the method includes: obtaining corresponding prediction basic data according to received event report data and performing binning processing to obtain basic feature data, screening model template parameters according to prediction configuration information in the event report data, and obtaining sample feature data corresponding to the event report data from a historical database to build a target prediction model, training the target prediction model according to the sample feature data to obtain model training information and perform attribution prediction on the basic feature data, and obtaining the predicted attribution population corresponding to each user in the event report data. The present invention belongs to the field of intelligent prediction technology, and can realize adaptive configuration of various scenarios and industries by obtaining model template parameters, obtain prediction basic data to realize multi-party data combination, solve the problem of inaccurate prediction results when using a general model for prediction, and meet the use requirements of multiple application fields.

Description

Technical Field

[0001] The present invention relates to the field of intelligent prediction technology, and in particular to a user attribution prediction method, device and equipment based on model adaptive adjustment. Background Art

[0002] In order to obtain the possible changes of users in the future based on their current behaviors, the user's belonging can be predicted, that is, to predict the user's future belonging group. The existing mainstream user prediction technology basically comes from the field of lookalike technology in the advertising field, which is mainly divided into two methods: explicit positioning and implicit positioning. However, the existing user prediction method is usually only applicable to a specific field. If the use field of the prediction method is expanded, the prediction method will be generalized, and accurate prediction results will not be obtained in a specific field. Therefore, the existing technical method is difficult to meet the differentiated use requirements between various application fields. Therefore, the user prediction method in the existing technical method cannot meet the differentiated use requirements between multiple application fields. Summary of the invention

[0003] The embodiments of the present invention provide a user attribution prediction method, device and equipment based on model adaptive adjustment, aiming to solve the problem that the user prediction method in the prior art cannot meet the differentiated usage requirements between various application fields.

[0004] In a first aspect, an embodiment of the present invention provides a user attribution prediction method based on model adaptive adjustment, the method is applied to a management server, a network connection is established between the management server and a client to realize data information transmission, the method includes:

[0005] If event reporting data is received from the client, corresponding user data is obtained from a preset user database according to the event reporting data, and combined with the event reporting data to obtain corresponding prediction basic data;

[0006] The prediction basic data is binned according to a preset binning rule to obtain corresponding basic feature data;

[0007] According to the prediction configuration information of the event reporting data, a set of model template parameters corresponding to the prediction configuration information is obtained by screening from a preset prediction model set;

[0008] Acquire sample feature information corresponding to the event reporting data from a preset historical database;

[0009] Constructing a corresponding target prediction model according to the sample feature information and the model template parameters;

[0010] Training the target prediction model according to the sample feature information to obtain model training information;

[0011] The basic feature data is predicted to belong to according to the target prediction model and the model training information to obtain the predicted belonging group to which each user in the event reporting data belongs.

[0012] In a second aspect, an embodiment of the present invention provides a user attribution prediction device based on model adaptive adjustment, comprising:

[0013] A prediction basic information acquisition unit, configured to, if receiving event report data from the client, acquire corresponding user data from a preset user database according to the event report data, and combine the user data with the event report data to obtain corresponding prediction basic data;

[0014] A basic feature data acquisition unit, used to perform binning processing on the prediction basic data according to a preset binning rule to obtain corresponding basic feature data;

[0015] A model template parameter acquisition unit, configured to filter and obtain a set of model template parameters corresponding to the prediction configuration information from a preset prediction model set according to the prediction configuration information of the event reporting data;

[0016] A sample feature information acquisition unit, used to acquire sample feature information corresponding to the event reporting data from a preset historical database;

[0017] A target prediction model acquisition unit, used to construct a corresponding target prediction model according to the sample feature information and the model template parameters;

[0018] A model training information acquisition unit, used to train the target prediction model according to the sample feature information to obtain model training information;

[0019] The crowd belonging prediction unit is used to perform belonging prediction on the basic feature data according to the target prediction model and the model training information to obtain the predicted belonging crowd to which each user in the event reporting data belongs.

[0020] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the user attribution prediction method based on model adaptive adjustment described in the first aspect is implemented.

[0021] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the user attribution prediction method based on model adaptive adjustment described in the first aspect above.

[0022] The embodiments of the present invention provide a user attribution prediction method, device and equipment based on model adaptive adjustment. According to the received event report data, the corresponding prediction basic data is obtained and the basic feature data is obtained by binning. The model template parameters are obtained by screening according to the prediction configuration information in the event report data, and the sample feature data corresponding to the event report data is obtained from the historical database to build a target prediction model. The target prediction model is trained according to the sample feature data to obtain model training information and the attribution prediction is performed on the basic feature data to obtain the predicted attribution population corresponding to each user in the event report data. Through the above method, adaptive configuration of various scenarios and industries can be achieved by obtaining model template parameters, and the prediction basic data can be obtained to achieve multi-party data combination, which solves the problem of inaccurate prediction results when the general model is used for prediction due to differences in data distribution in different application fields, and meets the differentiated usage requirements between multiple application fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0024] Figure 1 A schematic diagram of a flow chart of a user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of an application scenario of a user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of a sub-process of a user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0027] Figure 4 A schematic diagram of another sub-process of the user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0028] Figure 5 A schematic diagram of another sub-process of the user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0029] Figure 6 A schematic diagram of another sub-process of the user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0030] Figure 7 A schematic diagram of the latter sub-process of the user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0031] Figure 8 A schematic diagram of a subsequent sub-process of the user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0032] Fig. 9 A schematic block diagram of a user attribution prediction device based on model adaptive adjustment provided by an embodiment of the present invention;

[0033] Fig.10 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0036] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0037] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0038] See also Figure 1 and Figure 2 , Figure 1 A schematic diagram of a flow chart of a user attribution prediction method based on model adaptive adjustment provided by an embodiment of the present invention; Figure 2Schematic diagram of application scenario of the user affiliation prediction method based on model adaptive adjustment provided in an embodiment of the present invention; the user affiliation prediction method based on model adaptive adjustment is applied to a management server 10, and the user affiliation prediction method based on model adaptive adjustment is executed by application software installed in the management server 10, and a network connection is established between the management server 10 and the client 20 to realize the transmission of data information; the management server 10 is a server end for executing the user affiliation prediction method based on model adaptive adjustment to realize intelligent prediction of user affiliation, and the client 20 is a terminal device for communicating with the management server 10, such as a desktop computer, a laptop computer, a tablet computer or a mobile phone. Figure 2 The figure only illustrates the establishment of a network connection between one client 20 and the management server 10. In actual application, the management server 10 can simultaneously establish network connections with multiple clients 20 to achieve data information transmission.

[0039] like Figure 1 As shown, the method includes steps S110 to S170.

[0040] S110: If event reporting data is received from the client, corresponding user data is obtained from a preset user database according to the event reporting data, and combined with the event reporting data to obtain corresponding prediction basic data.

[0041] If the event reporting data is received from the client, the corresponding user data is obtained from the preset user database according to the event reporting data, and combined with the event reporting data to obtain the corresponding prediction basic data. Specifically, the client can report the obtained event reporting data to the management server through the SDK (Software Development Kit). In the process of using the application software, the user will generate interactive events with the background server. The event reporting data may include the behavior data of one or more users. The user's behavior data is the data generated by various usage behaviors during the user's use of the application. By obtaining the behavior data of each user, the user's usage habits of the corresponding application can be understood. The event reporting data includes more general data such as the user's historical active time, the number of historical clicks on push messages, etc., as well as data unique to e-commerce applications such as collections and purchases, or data unique to short video applications such as likes, collections, forwarding, and rewards.

[0042] The client can send the event reporting data obtained by the background server corresponding to the application to the management server. The management server is pre-configured with a user database, which can be used to record user-related data information, such as user hardware information, user gender, age, region and other data information. The user data corresponding to the corresponding user can be obtained from the user database according to the user identifier in the event reporting data. Each user identifier in the event reporting data corresponds to a piece of user data in the user database, so the user data corresponding to each user identifier in the event reporting data can be obtained, and the event reporting data can be combined with the corresponding user data to obtain the prediction basic data.

[0043] S120: binning the prediction basic data according to preset binning rules to obtain corresponding basic feature data.

[0044] The predicted basic data is binned according to the preset binning rules to obtain the corresponding basic feature data. After obtaining the user basic predicted data, the basic data of each user can be binned to obtain the basic feature data corresponding to the predicted basic data.

[0045] In one embodiment, if Figure 3 As shown, step S120 includes sub-steps S121, S122, S123 and S124.

[0046] S121. Count the behavior data of each user in the prediction basic data according to the statistical period in the binning rule to obtain corresponding behavior period statistical information.

[0047] Specifically, different types of data in the prediction basic data can be processed separately, and the behavior data of each user can be counted according to the statistical period in the binning rules. The user's behavior data is time-series information, so the user's behavior data over a period of time can be counted according to a fixed statistical period, such as the number of active days, active frequency, and number of push message clicks of users in the past month, so as to obtain the number of active days, active frequency, and push message clicks of users in each week within a month. Through the above method, the behavior data of each user can be counted separately to obtain the corresponding behavior cycle statistical information.

[0048] S122: convert the user basic information of each user in the prediction basic data according to the value conversion table in the binning rule to obtain corresponding basic value information.

[0049] The prediction basic data also includes the basic user information of each user, which includes hardware information, user gender, age, region, user preferences and other non-time-series information. The information corresponding to each indicator in the basic user information can be converted separately through the numerical conversion table, so as to convert the non-time-series information into basic numerical information represented by numerical values, that is, to convert the discrete feature information into one-hot encoding. For example, if the hardware information is R5-Android, the basic numerical information converted according to the numerical conversion table is 1035.

[0050] S123, performing chi-square binning processing on the basic numerical information according to the chi-square calculation formula and the chi-square threshold in the binning rule to obtain corresponding binning feature data.

[0051] The binning rules also include the chi-square calculation formula and the chi-square threshold, which can be used to perform chi-square binning on the information corresponding to a certain indicator in the basic numerical information, that is, to obtain the data values ​​corresponding to a certain indicator information and all users and perform binning. Specifically, first divide the data values ​​corresponding to a certain indicator information into several bins, each bin contains several data values, and calculate the chi-square values ​​of two adjacent bins according to the chi-square calculation formula, and determine whether the chi-square values ​​of the two adjacent bins are greater than the chi-square threshold. If greater, no processing is performed; if not, the adjacent bins with low chi-square values ​​are merged. Specifically, the chi-square calculation formula can be expressed by formula (1):

[0052]

[0053] Among them, x 2 is the calculated chi-square value, k is the number of categorical variable types, m = 2 (only the chi-square values ​​of two adjacent boxes are calculated), A ij That is, the observation frequency of the jth data value in the i-th box, E ij That is, ij The corresponding expected frequency.

[0054] After the chi-square binning process, the data value corresponding to each indicator information can be divided into different bins obtained after the binning process. Different bins are independent of each other, and the number corresponding to the bin to which the data value belongs is used as the binning feature data corresponding to the data value. Through this method, the binning feature data corresponding to each user can be obtained from the basic numerical information.

[0055] For example, the data value corresponding to the hardware information is 1035. After the data value of this indicator information corresponding to the hardware information is subjected to chi-square binning processing, 1035 belongs to the third bin, and "3" is used as the binning feature data corresponding to the data value "1035".

[0056] S124: normalize the periodic statistical information and combine it with the binned feature data to obtain basic feature data corresponding to each user in the prediction basic data.

[0057] Specifically, the logarithm of each data value in the periodic statistical information can be taken for normalization, and the normalized feature data of the same user can be combined with the binned feature data to obtain the basic feature data corresponding to the user. The basic feature data corresponding to each user can be obtained in turn through the above method.

[0058] S130. Filtering from a preset prediction model set according to the prediction configuration information of the event reporting data to obtain a set of model template parameters corresponding to the prediction configuration information.

[0059] According to the prediction configuration information of the event reporting data, a set of model template parameters corresponding to the prediction configuration information is obtained by screening from the preset prediction model set. The event reporting data also includes the prediction configuration information, and the prediction configuration information includes the industry type and the scenario type to which the application belongs. The prediction model set includes multiple configured models, and the parameters of a matching model can be obtained from the prediction model set based on the industry type and scenario type included in the prediction configuration information as the model template parameters.

[0060] In one embodiment, if Figure 4 As shown, step S130 includes sub-steps S131 and S132.

[0061] S131, according to the industry type in the prediction configuration information, filter out a corresponding first candidate model from the prediction model set. S132, according to the scenario type in the prediction configuration information, filter out a corresponding target model from the first candidate model, and obtain parameters of the target model as corresponding model template parameters.

[0062] Specifically, the first candidate model matching the industry type can be screened from the prediction model set according to the industry type. If the first candidate model matching the industry type is empty, a general template in the prediction model set is selected for use, such as the industry type can be the catering industry, taxi industry, shopping industry, etc. A target model matching the scene type is screened from the first candidate model according to the scene type, and the parameters in the target model are obtained as model template parameters. For example, for the catering industry, the scene type can be in-store ordering, takeaway delivery, fresh food delivery, etc. If the target model matching the scene type is empty, a general template in the prediction model set is selected for use. In addition, before performing template screening, it can also be determined whether the industry type and scene type in the prediction configuration information are empty. If the industry type or scene type is empty, a general template can also be selected for use. In the embodiment of this scheme, abstract industry and scene templates can be used to adapt to the data distribution differences of different industry types and scene types.

[0063] S140: Acquire sample feature information corresponding to the event reporting data from a preset historical database.

[0064] The sample characteristic information corresponding to the event report data is obtained from a preset historical database. The historical database stores historical data corresponding to the event report data. Then, according to the user identifier in the event report data, the corresponding characteristic information of the user can be obtained from the historical database as the sample characteristic information, that is, the characteristic information of each user in the historical database can be used as a sample.

[0065] In one embodiment, if Figure 5 As shown, step S140 includes sub-steps S141 and S142.

[0066] S141. Acquire historical sample data corresponding to the event reporting data from the historical database.

[0067] The historical sample data corresponding to each user ID can be obtained from the historical database according to the user ID contained in the event report data. The historical sample data includes the user's behavior data and basic user information over a period of time. For example, the user basic data of the behavior data set from two months ago to one month ago in the historical sample data is selected as the historical sample data.

[0068] S142: bin the historical sample data according to the binning rule to obtain corresponding sample feature information.

[0069] The historical sample data is binned according to the binning rules. Specifically, the binning results of the chi-square binning process can be selected to bin the user basic information in the historical sample data, and the corresponding number of the box that each data value in the historical sample data falls into can be directly obtained. After that, the behavioral data in the historical sample data is counted according to the statistical period and normalized by taking the logarithm, and the processing results can be combined to obtain the corresponding sample feature information.

[0070] S150: construct a corresponding target prediction model according to the sample feature information and the model template parameters.

[0071] The corresponding target prediction model is constructed according to the sample feature information and the model template parameters. The target prediction model can be constructed according to the sample feature information and the model template parameters. In order to achieve accurate prediction, multiple target prediction models can be constructed, that is, the construction of a multi-channel model is realized, and comprehensive prediction is performed through the constructed multiple target prediction models, thereby improving the accuracy of user attribution prediction.

[0072] In one embodiment, if Figure 6 As shown, step S150 includes sub-steps S151, S152 and S153.

[0073] S151. Group the samples included in the sample feature information to obtain a plurality of sample feature groups.

[0074] The samples included in the sample feature information can be grouped to obtain multiple sample feature groups, each of which contains data values ​​corresponding to multiple samples. The number of groups can be pre-configured, such as 50 groups, which means that the samples included in the sample feature information can be evenly divided into 50 sample feature groups.

[0075] S152, sorting the indicators in ascending and descending order according to the data amount of each indicator corresponding to the model template parameter in each sample feature group, to obtain an indicator sorting result corresponding to each sample feature group.

[0076] The model template parameters include multiple indicators for building the model. The data volume of each indicator corresponding to the model template parameters in each sample feature group can be obtained. The indicators can be sorted in ascending and descending order according to the data volume. The specific formula can be expressed by formula (2):

[0077] s=∑1 / rank i (2);

[0078] Among them, rank i Rank the data volume of the i-th indicator in the sample feature group, and sort the indicators in ascending or descending order according to the calculated correlation.

[0079] S153. Construct multiple target prediction models based on binary classification corresponding to each of the sample feature groups according to the indicator ranking results and the current population to which the samples in each of the sample feature groups belong.

[0080] The sample feature group also records the current population corresponding to each sample. The current population is the result of classifying the population to which the sample currently belongs based on the current classification standard. The indicator ranking results of each sample feature group and the current population of the samples in the sample feature group can be combined to construct a target prediction model corresponding to each sample feature group. Then, one sample feature group can correspond to one target prediction model. The constructed target prediction models are all prediction models based on binary classification problems.

[0081] Specifically, the data of the samples in the sample feature group can be used as input, and the current population corresponding to the samples can be used as the prediction result, and converted into a binary classification problem, so as to construct a prediction model based on the binary classification problem.

[0082] S160: Train the target prediction model according to the sample feature information to obtain model training information.

[0083] The target prediction model is trained according to the sample feature information to obtain model training information. The obtained multiple target prediction models can be trained according to the sample feature information, and the model training information of each target prediction model can be obtained based on the model training result.

[0084] In one embodiment, if Figure 7 As shown, step S160 includes sub-steps S161, S162 and S163.

[0085] S161. Extract a test sample set from the sample feature information.

[0086] First, a specific number of samples can be extracted from the sample feature information as a test sample set. For example, if the specific number is 50, 50 samples can be randomly extracted from the sample feature information to form a test sample set.

[0087] S162: Input the test sample set into each of the target prediction models for training, and obtain a training prediction result corresponding to each sample in the test sample set.

[0088] The characteristic information corresponding to the samples contained in the test sample set is input into each target prediction model for training. Each target prediction model can predict the characteristic information of the same sample and obtain the corresponding training prediction result. The training prediction result is the population to which the sample belongs during the training process of the target prediction model.

[0089] S163. Calculate the accuracy and recall rate of each target prediction model according to the training prediction results as corresponding model training information.

[0090] The training prediction results of each target prediction model on the test sample set are obtained, and the accuracy and recall of each target prediction model are calculated based on the training prediction results. The accuracy of the target prediction model is the ratio of the number of samples whose training prediction results are the same as the current belonging population to the total number of samples in the test sample set. The calculation formula of the recall rate can be expressed by formula (3):

[0091]

[0092] Among them, P r is the number of samples of the rth population in the training prediction results of the target prediction model, Q r is the number of samples of the rth population in the test sample set, R is the total number of population categories, and z is the recall rate of the corresponding target prediction model calculated.

[0093] S170. Perform attribution prediction on the basic feature data according to the target prediction model and the model training information to obtain the predicted attribution group to which each user in the event reporting data belongs.

[0094] The basic feature data is predicted to belong to according to the target prediction model and the model training information, so as to obtain the predicted belonging group to which each user in the event reporting data belongs. The belonging prediction can be performed on each user in the basic feature data according to the obtained model training information and the target prediction model, so as to obtain the predicted belonging group to which each user belongs.

[0095] In one embodiment, if Figure 8 As shown, step S170 includes sub-steps S171, S172 and S173.

[0096] S171. Perform attribution prediction on the basic feature data according to each of the target prediction models to obtain an initial attribution prediction result corresponding to each user in the basic feature data.

[0097] The characteristic data of each user in the basic characteristic data can be input into the target prediction model for attribution prediction, so as to obtain the initial attribution prediction result corresponding to each user. Then the number of initial attribution prediction results of each user is equal to the number of target prediction models.

[0098] S172. Determine a division point according to the accuracy and recall rate corresponding to each target prediction model in the model training information.

[0099] The target prediction models in the model training information can be sorted according to their accuracy, and the intersection of the accuracy and recall curves can be selected as the dividing point.

[0100] S173. According to the division point, obtain the prediction result higher than the division point from the initial attribution prediction results as the predicted attribution population of each user.

[0101] According to the division point, the prediction result is obtained from the initial attribution prediction result as the user's predicted attribution population. Specifically, the prediction results of the target prediction model with higher accuracy and recall than the division point for all users can be obtained. Since there may be multiple target prediction models higher than the division point, the population with the largest number of predictions is selected according to the prediction result of a user as the user's predicted attribution population. Obtaining the predicted attribution population means to determine the population to which the user is most likely to belong in the future (e.g., one month later) based on the current prediction basic data.

[0102] In the user attribution prediction method based on model adaptive adjustment provided in the embodiment of the present invention, the corresponding prediction basic data is obtained according to the received event report data and the basic feature data is obtained by binning, the model template parameters are obtained according to the prediction configuration information in the event report data, and the sample feature data corresponding to the event report data is obtained from the historical database to build a target prediction model, the target prediction model is trained according to the sample feature data to obtain model training information and the attribution prediction is performed on the basic feature data to obtain the predicted attribution population corresponding to each user in the event report data. Through the above method, adaptive configuration of various scenarios and industries can be achieved by obtaining model template parameters, and the prediction basic data can be obtained to achieve multi-party data combination, which solves the problem of inaccurate prediction results when the general model is used for prediction due to differences in data distribution in different application fields, and meets the differentiated usage requirements between various application fields.

[0103] The embodiment of the present invention further provides a user attribution prediction device based on model adaptive adjustment, which can be configured in the management server 10, and is used to execute any embodiment of the user attribution prediction method based on model adaptive adjustment. Fig. 9 , Fig. 9 A schematic block diagram of a user attribution prediction device based on model adaptive adjustment provided by an embodiment of the present invention.

[0104] like Fig. 9As shown, the user affiliation prediction device 100 based on model adaptive adjustment includes a prediction basic information acquisition unit 110, a basic feature data acquisition unit 120, a model template parameter acquisition unit 130, a sample feature information acquisition unit 140, a target prediction model acquisition unit 150, a model training information acquisition unit 160 and a population affiliation prediction unit 170.

[0105] The prediction basic information acquisition unit 110 is used to, if receiving event reporting data from the client, acquire corresponding user data from a preset user database according to the event reporting data, and combine it with the event reporting data to obtain corresponding prediction basic data.

[0106] The basic feature data acquisition unit 120 is used to perform binning processing on the prediction basic data according to a preset binning rule to obtain corresponding basic feature data.

[0107] In a specific embodiment, the basic feature data acquisition unit 120 includes subunits: a period statistical information acquisition unit, which is used to perform statistics on the behavior data of each user in the predicted basic data according to the statistical period in the binning rule to obtain corresponding behavior period statistical information; a basic numerical information acquisition unit, which is used to convert the user basic information of each user in the predicted basic data according to the numerical conversion table in the binning rule to obtain corresponding basic numerical information; a chi-square binning processing unit, which is used to perform chi-square binning processing on the basic numerical information according to the chi-square calculation formula and chi-square threshold in the binning rule to obtain corresponding binning feature data; a feature combination unit, which is used to normalize the period statistical information and then combine it with the binning feature data to obtain basic feature data corresponding to each user in the predicted basic data.

[0108] The model template parameter acquisition unit 130 is used to filter out a set of model template parameters corresponding to the prediction configuration information from a preset prediction model set according to the prediction configuration information of the event reporting data.

[0109] In a specific embodiment, the model template parameter acquisition unit 130 includes sub-units: a first model screening unit, used to filter out a corresponding first candidate model from the prediction model set according to the industry type in the prediction configuration information; a second model screening unit, used to filter out a corresponding target model from the first candidate model according to the scenario type in the prediction configuration information, and obtain the parameters of the target model as the corresponding model template parameters.

[0110] The sample characteristic information acquisition unit 140 is used to acquire the sample characteristic information corresponding to the event reporting data from a preset history database.

[0111] In a specific embodiment, the sample feature information acquisition unit 140 includes sub-units: a historical sample data acquisition unit, used to obtain historical sample data corresponding to the event reporting data from the historical database; and a data binning processing unit, used to bin the historical sample data according to the binning rules to obtain corresponding sample feature information.

[0112] The target prediction model acquisition unit 150 is used to construct a corresponding target prediction model according to the sample feature information and the model template parameters.

[0113] In a specific embodiment, the target prediction model acquisition unit 150 includes subunits: a sample grouping unit, which is used to group the samples contained in the sample feature information to obtain multiple sample feature groups; an indicator sorting result acquisition unit, which is used to sort the indicators in ascending and descending order according to the data volume of each indicator corresponding to the model template parameter in each sample feature group, and obtain the indicator sorting result corresponding to each sample feature group; a model construction unit, which is used to construct multiple target prediction models based on binary classification corresponding to each sample feature group according to the indicator sorting result and the current population to which the samples in each sample feature group belong.

[0114] The model training information acquisition unit 160 is used to train the target prediction model according to the sample feature information to obtain model training information.

[0115] In a specific embodiment, the model training information acquisition unit 160 includes subunits: a test sample set acquisition unit, which is used to extract a test sample set from the sample feature information; a training prediction result acquisition unit, which is used to input the test sample set into each of the target prediction models for training, and obtain the training prediction result corresponding to each sample in the test sample set; and a calculation unit, which is used to calculate the accuracy and recall rate of each of the target prediction models according to the training prediction results as the corresponding model training information.

[0116] The group belonging prediction unit 170 is used to perform belonging prediction on the basic feature data according to the target prediction model and the model training information to obtain the predicted belonging group to which each user in the event reporting data belongs.

[0117] In a specific embodiment, the population attribution prediction unit 170 includes subunits: an initial attribution prediction result acquisition unit, which is used to perform attribution prediction on the basic feature data according to each of the target prediction models, and obtain the initial attribution prediction result corresponding to each user in the basic feature data; a division point determination unit, which is used to determine the division point according to the accuracy and recall rate corresponding to each of the target prediction models in the model training information; and a predicted attribution population acquisition unit, which is used to obtain the prediction result higher than the division point from the initial attribution prediction result according to the division point as the predicted attribution population of each user.

[0118] The user attribution prediction device based on model adaptive adjustment provided in the embodiment of the present invention applies the above-mentioned user attribution prediction method based on model adaptive adjustment, obtains the corresponding prediction basic data according to the received event report data and performs binning processing to obtain basic feature data, obtains model template parameters according to the prediction configuration information in the event report data, obtains sample feature data corresponding to the event report data from the historical database to build a target prediction model, trains the target prediction model according to the sample feature data to obtain model training information and performs attribution prediction on the basic feature data, and obtains the predicted attribution population corresponding to each user in the event report data. Through the above method, adaptive configuration of various scenarios and industries can be achieved by obtaining model template parameters, and the prediction basic data can be obtained to achieve multi-party data combination, which solves the problem of inaccurate prediction results when the general model is used for prediction due to differences in data distribution in different application fields, and meets the differentiated usage requirements between various application fields.

[0119] The user attribution prediction device based on model adaptive adjustment can be implemented in the form of a computer program. The computer program can be used in Fig.10 Runs on the computer device shown.

[0120] See also Fig.10 , Fig.10 1 is a schematic block diagram of a computer device provided by an embodiment of the present invention. The computer device may be a management server 10 for executing a user attribution prediction method based on model adaptive adjustment to achieve intelligent prediction of user attribution.

[0121] See also Fig.10 The computer device 500 includes a processor 502 , a memory and a network interface 505 connected via a system bus 501 , wherein the memory may include a storage medium 503 and an internal memory 504 .

[0122] The storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, the processor 502 may execute a user attribution prediction method based on model adaptive adjustment, wherein the storage medium 503 may be a volatile storage medium or a non-volatile storage medium.

[0123] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0124] The internal memory 504 provides an environment for the operation of the computer program 5032 in the storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute the user attribution prediction method based on model adaptive adjustment.

[0125] The network interface 505 is used for network communication, such as providing data information transmission, etc. Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device 500 to which the solution of the present invention is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0126] The processor 502 is used to run the computer program 5032 stored in the memory to implement the corresponding functions in the above-mentioned user attribution prediction method based on model adaptive adjustment.

[0127] Those skilled in the art will understand that Fig.10 The embodiments of the computer device shown in the figure do not constitute a limitation on the specific composition of the computer device. In other embodiments, the computer device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, in some embodiments, the computer device may only include a memory and a processor. In such embodiments, the structure and function of the memory and the processor are the same as those of the embodiment of the present invention. Fig.10 The embodiments shown are consistent and will not be described again here.

[0128] It should be understood that in the embodiment of the present invention, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0129] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium may be a volatile or non-volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps included in the user attribution prediction method based on model adaptive adjustment are implemented.

[0130] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the above-described equipment, devices and units can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. Those of ordinary skill in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0131] In the several embodiments provided by the present invention, it should be understood that the disclosed equipment, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. Units with the same function may also be combined into one unit. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices or units, or may be an electrical, mechanical or other form of connection.

[0132] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0133] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0134] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a computer-readable storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a magnetic disk or an optical disk.

[0135] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A user attribution prediction method based on model adaptive adjustment, characterized in that: The method is applied to a management server, and a network connection is established between the management server and a client to realize the transmission of data information. The method includes: If event reporting data is received from the client, corresponding user data is obtained from a preset user database according to the event reporting data, and combined with the event reporting data to obtain corresponding prediction basic data; the event reporting data includes behavior data of one or more users, and the behavior data is data generated by the user's use behavior during the use of the application program; The prediction basic data is binned according to a preset binning rule to obtain corresponding basic feature data; According to the prediction configuration information of the event reporting data, a set of model template parameters corresponding to the prediction configuration information is obtained by screening from a preset prediction model set; Acquire sample feature information corresponding to the event reporting data from a preset historical database; Constructing a corresponding target prediction model according to the sample feature information and the model template parameters; Training the target prediction model according to the sample feature information to obtain model training information; Performing attribution prediction on the basic feature data according to the target prediction model and the model training information to obtain the predicted attribution group to which each user in the event reporting data belongs; The method of selecting a set of model template parameters corresponding to the prediction configuration information from a preset prediction model set according to the prediction configuration information of the event reporting data includes: According to the industry type in the forecast configuration information, a corresponding first candidate model is obtained by screening from the forecast model set; A corresponding target model is screened out from the first candidate models according to the scenario type in the prediction configuration information, and parameters of the target model are obtained as corresponding model template parameters.

2. The user attribution prediction method based on model adaptive adjustment according to claim 1, characterized in that: The binning process of the prediction basic data according to the preset binning rules to obtain corresponding basic feature data includes: According to the statistical period in the binning rule, statistics are collected on the behavior data of each user in the prediction basic data to obtain corresponding behavior period statistical information; According to the value conversion table in the binning rule, the user basic information of each user in the prediction basic data is converted to obtain corresponding basic numerical information; Perform chi-square binning processing on the basic numerical information according to the chi-square calculation formula and chi-square threshold in the binning rule to obtain corresponding binning feature data; The periodic statistical information is normalized and then combined with the binned feature data to obtain basic feature data corresponding to each user in the prediction basic data.

3. The user attribution prediction method based on model adaptive adjustment according to claim 1, characterized in that: The acquiring of sample feature information corresponding to the event reporting data from a preset historical database includes: Acquire historical sample data corresponding to the event reporting data from the historical database; The historical sample data is binned according to the binning rules to obtain corresponding sample feature information.

4. The user attribution prediction method based on model adaptive adjustment according to claim 1, characterized in that: The constructing a corresponding target prediction model according to the sample feature information and the model template parameters includes: Grouping the samples included in the sample feature information to obtain a plurality of sample feature groups; According to the data amount of each indicator corresponding to the model template parameter in each sample feature group, the indicators are sorted in ascending and descending order to obtain the indicator sorting result corresponding to each sample feature group; According to the index ranking results and the current population to which the samples in each sample feature group belong, multiple target prediction models based on binary classification corresponding to each sample feature group are constructed.

5. The user attribution prediction method based on model adaptive adjustment according to claim 1, characterized in that: The step of training the target prediction model according to the sample feature information to obtain model training information includes: Extracting a test sample set from the sample feature information; Inputting the test sample set into each of the target prediction models for training respectively, and obtaining a training prediction result corresponding to each sample in the test sample set; The accuracy and recall of each target prediction model are calculated according to the training prediction results as corresponding model training information.

6. The user attribution prediction method based on model adaptive adjustment according to claim 1, characterized in that: The performing attribution prediction on the basic feature data according to the target prediction model and the model training information to obtain the predicted attribution group to which each user in the event reporting data belongs includes: According to each of the target prediction models, the basic feature data is respectively predicted to obtain an initial prediction result of the basic feature data corresponding to each user; Determine a division point according to the accuracy and recall rate corresponding to each of the target prediction models in the model training information; According to the division point, a prediction result higher than the division point is obtained from the initial attribution prediction result as the predicted attribution population of each user.

7. A user attribution prediction device based on model adaptive adjustment, characterized in that: The device is configured in a management server, and a network connection is established between the management server and the client to realize the transmission of data information. The device includes: A prediction basic information acquisition unit, configured to, if receiving event report data from the client, acquire corresponding user data from a preset user database according to the event report data, and combine the data with the event report data to obtain corresponding prediction basic data; the event report data includes behavior data of one or more users, and the behavior data is data generated by the user's use behavior during the use of the application program; A basic feature data acquisition unit, used to perform binning processing on the prediction basic data according to a preset binning rule to obtain corresponding basic feature data; A model template parameter acquisition unit, configured to filter and obtain a set of model template parameters corresponding to the prediction configuration information from a preset prediction model set according to the prediction configuration information of the event reporting data; A sample feature information acquisition unit, used to acquire sample feature information corresponding to the event reporting data from a preset historical database; A target prediction model acquisition unit, used to construct a corresponding target prediction model according to the sample feature information and the model template parameters; A model training information acquisition unit, used to train the target prediction model according to the sample feature information to obtain model training information; A crowd belonging prediction unit, used to perform belonging prediction on the basic feature data according to the target prediction model and the model training information, so as to obtain the predicted belonging crowd to which each user in the event reporting data belongs; The method of selecting a set of model template parameters corresponding to the prediction configuration information from a preset prediction model set according to the prediction configuration information of the event reporting data includes: According to the industry type in the forecast configuration information, a corresponding first candidate model is obtained by screening from the forecast model set; A corresponding target model is screened out from the first candidate models according to the scenario type in the prediction configuration information, and parameters of the target model are obtained as corresponding model template parameters.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the user attribution prediction method based on model adaptive adjustment according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the user attribution prediction method based on model adaptive adjustment according to any one of claims 1 to 6 is implemented.

Images