CN113190599B

CN113190599B - Processing method, device and equipment for application user behavior data and storage medium

Info

Publication number: CN113190599B
Application number: CN202110735749.0A
Authority: CN
Inventors: 沈嘉良; 胡英东; 李雪丽; 贾素苇; 徐宁; 陶醉; 陈煦
Original assignee: Ping An Technology Shenzhen Co Ltd
Current assignee: Ping An Technology Shenzhen Co Ltd
Priority date: 2021-06-30
Filing date: 2021-06-30
Publication date: 2021-09-28
Anticipated expiration: 2041-06-30
Also published as: CN113190599A; WO2023273299A1

Abstract

The invention relates to the field of artificial intelligence, and provides a method, a device, equipment and a storage medium for processing application user behavior data, which are used for improving the accuracy of a business operation adjustment strategy based on application user behavior data analysis. The processing method of the application user behavior data comprises the following steps: performing preset type classification, sequence data conversion and classification statistics on the application user behavior data to obtain an application operation recording sequence and application data to be processed; predicting subsequent business operation intention of the application operation record sequence to obtain business operation intention prediction data; classifying the application operation recording sequence to obtain dynamic score data, and classifying the application data to be processed to obtain static score data; and acquiring a business operation adjustment strategy to be pushed according to the dynamic score data, the static score data, the business operation intention prediction data and the target score data. In addition, the invention also relates to a block chain technology, and the application user behavior data can be stored in the block chain.

Description

Processing method, device and equipment for application user behavior data and storage medium

Technical Field

The invention relates to the field of intelligent decision making of artificial intelligence, in particular to a method, a device, equipment and a storage medium for processing application user behavior data.

Background

With the development of internet technology and computer technology, users gradually implement various corresponding business operations through various operation applications, in order to be able to timely and effectively push the applications to the users, collect static data of application programs, predict the business operation intention of the users based on data analysis indexes (such as daily active user number DAU, monthly active user number MAU, newly added user number, user retention rate, jump rate, loss rate and transaction rate among pages, etc.) of the application programs, obtain a prediction result, and send the business operation adjustment strategy corresponding to the prediction result to the user side.

In the method, the adopted data analysis indexes are mechanized-based data indexes, which only can reflect the overall use condition of the user and cannot perform deep analysis on the operation behavior of a single user, so that the prediction accuracy of the business operation intention is low, and the accuracy of the business operation adjustment strategy based on the application user behavior data analysis is low.

Disclosure of Invention

The invention provides a method, a device, equipment and a storage medium for processing application user behavior data, which are used for improving the accuracy of a business operation adjustment strategy based on application user behavior data analysis.

The invention provides a processing method of application user behavior data in a first aspect, which comprises the following steps:

acquiring application user behavior data, and performing preset type classification, sequence data conversion and classification statistics on the application user behavior data to obtain an application operation record sequence and application data to be processed;

predicting subsequent business operation intention of the application operation record sequence through a preset cyclic neural network model to obtain business operation intention prediction data;

calling a preset dynamic classification model, classifying the application operation record sequence based on business operation intention to obtain dynamic score data, calling a preset static classification model, and classifying the application data to be processed based on business operation intention to obtain static score data;

and summing the dynamic score data and the static score data to obtain target score data, and acquiring a business operation adjustment strategy to be pushed from a preset adjustment strategy sequence according to the business operation intention prediction data and the target score data.

Optionally, in a first implementation manner of the first aspect of the present invention, the acquiring application user behavior data, and performing preset type classification, sequence data conversion, and classification statistics on the application user behavior data to obtain an application operation recording sequence and application data to be processed includes:

acquiring application user behavior data, and classifying the application user behavior data according to a service operation result and a preset type to obtain initial application operation data and initial application data, wherein the preset type comprises an application operation data type and an application user data type;

performing linked list storage based on a preset time period on the initial application operation data to obtain an application operation record sequence;

and carrying out operation object classification and record statistics based on preset dimensionality on the initial application data to obtain application data to be processed.

Optionally, in a second implementation manner of the first aspect of the present invention, the predicting subsequent business operation intention of the application operation record sequence by using a preset recurrent neural network model to obtain business operation intention prediction data includes:

performing subsequent business operation prediction based on a business operation result on the application operation record sequence through a preset recurrent neural network model to obtain target subsequent business operation information, wherein the target subsequent business operation information comprises target subsequent business operation data and a probability value of the target subsequent business operation data based on the business operation result;

acquiring subsequent operation data of the application user behavior data, and matching the subsequent operation data of the application user behavior data with the target subsequent service operation data to obtain matched target subsequent service operation data;

and determining the probability value corresponding to the matched target subsequent business operation data as business operation intention prediction data.

Optionally, in a third implementation manner of the first aspect of the present invention, the invoking a preset dynamic classification model to classify the application operation record sequence based on business operation intention to obtain dynamic score data, and invoking a preset static classification model to classify the application data to be processed based on business operation intention to obtain static score data includes:

calling a first classification model in a preset dynamic classification model, and extracting attention characteristics of the application operation record sequence, calculating probability value based on business operation intention and judging the probability value to obtain attention score data;

performing feature extraction and classification on the application operation record sequence through a second classification model in the dynamic classification model to obtain original score data;

summing the attention score data and the original score data to obtain dynamic score data;

and calling a preset static classification model, and performing feature extraction, attention feature fusion, probability value calculation and probability value discrimination based on business operation intention on the application data to be processed to obtain static score data.

Optionally, in a fourth implementation manner of the first aspect of the present invention, the summing the dynamic score data and the static score data to obtain target score data, and acquiring a to-be-pushed business operation adjustment policy from a preset adjustment policy sequence according to the business operation intention prediction data and the target score data, includes:

calculating the sum of the dynamic score data and the static score data according to preset weight to obtain target score data;

comparing and analyzing the target score data with the business operation intention prediction data and a preset threshold respectively to obtain an intention analysis result and a threshold analysis result;

if the intention analysis result is that the service operation intention prediction data is smaller than the target score data and/or the threshold analysis result is that the target score data is smaller than a preset threshold, performing first adjustment strategy extraction on a preset adjustment strategy sequence to obtain a service operation adjustment strategy to be pushed, wherein the preset adjustment strategy sequence is an adjustment strategy set which performs reverse ordering according to the size of the score of the operation.

Optionally, in a fifth implementation manner of the first aspect of the present invention, if the intention analysis result indicates that the service operation intention prediction data is smaller than the target score data, and/or the threshold analysis result indicates that the target score data is smaller than a preset threshold, performing first adjustment policy extraction on a preset adjustment policy sequence, and after obtaining a service operation adjustment policy to be pushed, the method further includes:

and if the intention analysis result is that the business operation intention prediction data is larger than or equal to the target score data, performing reinforcement learning on the recurrent neural network model.

Optionally, in a sixth implementation manner of the first aspect of the present invention, before the acquiring application user behavior data, and performing preset type classification, sequence data conversion, and classification statistics on the application user behavior data to obtain an application operation recording sequence and application data to be processed, the method further includes:

acquiring historical operation sequence data, target user portrait data and recall configuration information of a user based on an application program to be processed, wherein the target user portrait data comprises a score of the operation data;

based on the historical operation sequence data and the recall configuration information, index retrieval and reading are carried out on the target user portrait data to obtain a score of the read operation data;

scoring the historical operation sequence data according to the score of the read operation data to obtain scored historical operation sequence data;

and according to the value of the score, performing descending arrangement on the scored historical operation sequence data to obtain a preset adjustment strategy sequence.

A second aspect of the present invention provides a processing apparatus for applying user behavior data, including:

the statistical module is used for acquiring application user behavior data, and performing preset type classification, sequence data conversion and classification statistics on the application user behavior data to obtain an application operation record sequence and application data to be processed;

the prediction module is used for predicting the subsequent business operation intention of the application operation record sequence through a preset cyclic neural network model to obtain business operation intention prediction data;

the classification module is used for calling a preset dynamic classification model, classifying the application operation record sequence based on business operation intention to obtain dynamic score data, calling a preset static classification model, and classifying the application data to be processed based on the business operation intention to obtain static score data;

and the first acquisition module is used for summing the dynamic score data and the static score data to obtain target score data, and acquiring a business operation adjustment strategy to be pushed from a preset adjustment strategy sequence according to the business operation intention prediction data and the target score data.

Optionally, in a first implementation manner of the second aspect of the present invention, the statistics module includes:

the classification unit is used for acquiring application user behavior data, classifying the application user behavior data according to a service operation result and a preset type to obtain initial application operation data and initial application data, wherein the preset type comprises an application operation data type and an application user data type;

the storage unit is used for storing the initial application operation data based on a linked list of a preset time period to obtain an application operation record sequence;

and the statistical unit is used for carrying out operation object classification and record statistics based on preset dimensionality on the initial application data to obtain application data to be processed.

Optionally, in a second implementation manner of the second aspect of the present invention, the prediction module is specifically configured to:

Optionally, in a third implementation manner of the second aspect of the present invention, the classification module is specifically configured to:

Optionally, in a fourth implementation manner of the second aspect of the present invention, the first obtaining module includes:

the calculating unit is used for calculating the sum of the dynamic score data and the static score data according to preset weight to obtain target score data;

the analysis unit is used for comparing and analyzing the target score data with the business operation intention prediction data and a preset threshold respectively to obtain an intention analysis result and a threshold analysis result;

and the extracting unit is used for extracting a first adjustment strategy from a preset adjustment strategy sequence to obtain a business operation adjustment strategy to be pushed if the intention analysis result indicates that the business operation intention prediction data is smaller than the target score data and/or the threshold analysis result indicates that the target score data is smaller than a preset threshold, wherein the preset adjustment strategy sequence is an adjustment strategy set which is sorted in a reverse order according to the scale values of the operations.

Optionally, in a fifth implementation manner of the second aspect of the present invention, the first obtaining module further includes:

and the reinforcement learning unit is used for performing reinforcement learning on the recurrent neural network model if the intention analysis result is that the business operation intention prediction data is greater than or equal to the target score data.

Optionally, in a sixth implementation manner of the second aspect of the present invention, the apparatus for processing application user behavior data further includes:

the second acquisition module is used for acquiring historical operation sequence data, target user portrait data and recall configuration information of a user based on the application program to be processed, wherein the target user portrait data comprises a score of the operation data;

the reading module is used for carrying out index retrieval and reading on the target user portrait data based on the historical operation sequence data and the recall configuration information to obtain a score of the read operation data;

the scoring module is used for scoring the historical operation sequence data according to the score of the read operation data to obtain scored historical operation sequence data;

and the sorting module is used for performing descending sorting on the scored historical operation sequence data according to the score value to obtain a preset adjustment strategy sequence.

A third aspect of the present invention provides a processing device for applying user behavior data, including: a memory and at least one processor, the memory having instructions stored therein; the at least one processor calls the instructions in the memory to enable the processing device applying the user behavior data to execute the processing method applying the user behavior data.

A fourth aspect of the present invention provides a computer-readable storage medium having stored therein instructions, which, when run on a computer, cause the computer to execute the above-described processing method of applying user behavior data.

According to the technical scheme provided by the invention, application user behavior data are obtained, and preset type classification, sequence data conversion and classification statistics are carried out on the application user behavior data to obtain an application operation record sequence and application data to be processed; predicting subsequent business operation intention of the application operation record sequence through a preset cyclic neural network model to obtain business operation intention prediction data; calling a preset dynamic classification model, classifying the application operation record sequence based on business operation intention to obtain dynamic score data, calling a preset static classification model, and classifying the application data to be processed based on the business operation intention to obtain static score data; and summing the dynamic score data and the static score data to obtain target score data, and acquiring a service operation adjustment strategy to be pushed from a preset adjustment strategy sequence according to the service operation intention prediction data and the target score data. In the embodiment of the invention, the application user behavior data is analyzed by combining the angles of machine learning and time series deep learning, the accuracy and comprehensiveness of the prediction result of the business operation intention are ensured, the operation behavior of a single user can be deeply analyzed, the classification of static data and dynamic time series data of the application user behavior data is realized, the prediction accuracy of the business operation intention is improved, and the accuracy of the business operation adjustment strategy based on the analysis of the application user behavior data is further improved.

Drawings

FIG. 1 is a schematic diagram of an embodiment of a processing method for applying user behavior data according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of another embodiment of a processing method for applying user behavior data according to an embodiment of the present invention;

FIG. 3 is a diagram of an embodiment of a processing device for applying user behavior data according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of another embodiment of a processing device for applying user behavior data according to an embodiment of the present invention;

fig. 5 is a schematic diagram of an embodiment of a processing device for applying user behavior data in the embodiment of the present invention.

Detailed Description

The embodiment of the invention provides a method, a device, equipment and a storage medium for processing application user behavior data, which improve the accuracy of a business operation adjustment strategy based on application user behavior data analysis.

The terms "first," "second," "third," "fourth," and the like in the description and in the claims, as well as in the drawings, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that the embodiments described herein may be practiced otherwise than as specifically illustrated or described herein. Furthermore, the terms "comprises," "comprising," or "having," and any variations thereof, are intended to cover non-exclusive inclusions, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

For convenience of understanding, a specific flow of the embodiment of the present invention is described below, and referring to fig. 1, an embodiment of a processing method for applying user behavior data in the embodiment of the present invention includes:

101. and acquiring application user behavior data, and performing preset type classification, sequence data conversion and classification statistics on the application user behavior data to obtain an application operation record sequence and application data to be processed.

It is to be understood that the execution subject of the present invention may be a processing device applying user behavior data, and may also be a terminal or a server, which is not limited herein. The embodiment of the present invention is described by taking a server as an execution subject.

After obtaining user authorization, a server calls a preset timer to execute a timing task, when a preset time is reached through the timing task, a preset crawler or a grabbing tool is called to grab application user behavior data of an application program (APP) to be processed, the number of the application programs to be processed comprises one or more than one, namely the application user behavior data can be data corresponding to one application program to be processed, the application user behavior data can also be data corresponding to more than one application program to be processed respectively, and the application user behavior data comprises recorded data generated by the application program to be processed based on user operation, user data based on the application program to be processed and state data of business operation of a user on the application program to be processed; missing value filling, same type combination and data desensitization are carried out on the application user behavior data to obtain the application user behavior data after pretreatment, and the quality and the safety of the application user behavior data are improved.

The server classifies the preprocessed application user behavior data according to preset types (the preset types comprise application operation data types and application user data types) to obtain the application operation data after pre-classification and the application data after pre-classification; performing data extraction on the application operation data after the pre-classification according to a preset time interval to obtain extracted application operation data; converting the extracted application operation data into time sequence data to obtain an application operation record sequence, wherein the application operation data after pre-classification comprises user application operation record data and user application purchase data; and performing functional area classification on the pre-classified application data to obtain classified user application data, and performing statistics on the classified user application data according to preset statistical indexes to obtain to-be-processed application data, wherein the preset statistical indexes include but are not limited to login duration, login times, operation times, residence time of each plate and preference labels of a user based on to-be-processed application programs.

102. And predicting subsequent business operation intention of the application operation record sequence through a preset cyclic neural network model to obtain business operation intention prediction data.

The server trains and optimizes the created recurrent neural network model to obtain a final recurrent neural network model (RNN) in advance through a preset historical application operation record sequence, wherein the recurrent neural network model is a deep bidirectional RNN and can predict subsequent optimal operation which is required to be performed when business operation is achieved in each current state. The server calls a preset recurrent neural network model, sequence feature extraction and activation function operation based on a memory state are carried out on an application operation recording sequence through an input layer, a plurality of hidden layers and an output layer in the recurrent neural network model to obtain target subsequent service operation information, the target subsequent service operation information comprises target subsequent service operation data and a probability value of the target subsequent service operation data based on a service operation result, and the target subsequent service operation data comprises optimal subsequent operation and suboptimal subsequent operation; and classifying the application operation record sequence based on the current business operation of each state according to the target subsequent business operation information to obtain business operation intention prediction data.

103. Calling a preset dynamic classification model, classifying the application operation record sequence based on business operation intention to obtain dynamic score data, calling a preset static classification model, and classifying the application data to be processed based on the business operation intention to obtain static score data.

The server trains and optimizes a preset dynamic prediction model by applying an operation record sequence sample in advance to obtain a dynamic classification model, wherein the dynamic classification model is used for classifying data which has a large operation record data amount and is a time sequence and has a certain dependency relationship with the previous overall operation in the subsequent operation, namely the dynamic classification model is used for classifying and predicting dynamic time sequence data (namely the application operation record sequence sample), and the dynamic classification model can be a long-short-term memory network (LSTM) model; the server trains and optimizes a preset static prediction model in advance through a user application data sample to obtain a static classification model, wherein the static classification model is used for classifying user service execution conditions (service operation intentions) of data counted at the same time, namely the static classification model is used for classifying and predicting static data (namely the user application data sample), and the static classification model can be extreme gradient boosting (xgboost), wherein the execution process of obtaining the application operation recording sequence sample and the user application data sample is similar to the execution process of the step 101, and is not repeated here.

The server calls a preset dynamic classification model, extracts multi-level time sequence features of the application operation recording sequence to obtain operation recording time sequence features, and calculates probability values and judges the probability values based on business operation intention of the operation recording time sequence features to obtain dynamic score data; calling a preset static classification model, extracting multi-level features of application data to be processed to obtain user application features, calculating probability values and judging the probability values based on business operation intentions of the user application features to obtain static score data, wherein the business operation intentions are as follows: the intention to purchase. The dynamic classification model and the static classification model can be a combination of a feature extraction network and a classification network, and the feature extraction network is used for performing multi-level feature extraction, for example: the feature extraction network can be a target detection model ET-Yoloov 3, and accuracy of dynamic score data and static score data is improved.

104. And summing the dynamic score data and the static score data to obtain target score data, and acquiring a service operation adjustment strategy to be pushed from a preset adjustment strategy sequence according to the service operation intention prediction data and the target score data.

The service operation adjustment policy includes, but is not limited to, an application adjustment policy, a same-type content adjustment policy, and a function adjustment policy of the service operation, for example: the business operation is used as a purchasing operation on the application program, in order to achieve purchasing or increase purchasing intention, the application program with higher user common use or score is pushed (namely, the application adjustment strategy), an operation path similar to the current operation in the application user behavior data is pushed (namely, the same type of content adjustment strategy), and a function module with similar business operation function corresponding to the application user behavior data is pushed (namely, the function adjustment strategy).

The server respectively performs two-dimensional matrix conversion on the dynamic score data and the static score data to obtain a dynamic two-dimensional matrix and a static two-dimensional matrix, and calculates the arithmetic mean of the dynamic two-dimensional matrix and the static two-dimensional matrix to obtain target score data; and respectively comparing and analyzing the target score data with the service operation intention prediction data and a preset threshold to obtain an intention analysis result and a threshold analysis result, if the intention analysis result is a result corresponding to pushing, obtaining a first strategy in descending order of the scores from a preset adjustment strategy sequence to obtain a service operation adjustment strategy to be pushed, or sending the intention analysis result and the threshold analysis result to a preset terminal, manually following through responsible personnel of the preset terminal, and if the intention analysis result is a result corresponding to which pushing is not required, not processing.

In the embodiment of the invention, the application user behavior data is analyzed by combining the angles of machine learning and time series deep learning, the accuracy and comprehensiveness of the prediction result of the business operation intention are ensured, the operation behavior of a single user can be deeply analyzed, the classification of static data and dynamic time series data of the application user behavior data is realized, the prediction accuracy of the business operation intention is improved, and the accuracy of the business operation adjustment strategy based on the analysis of the application user behavior data is further improved.

Referring to fig. 2, another embodiment of the processing method for applying user behavior data according to the embodiment of the present invention includes:

201. the method comprises the steps of obtaining application user behavior data, classifying the application user behavior data according to a service operation result and a preset type to obtain initial application operation data and initial application data, wherein the preset type comprises an application operation data type and an application user data type.

After obtaining user authorization, the server regularly collects application user behavior data, classifies the application user behavior data according to a service operation result, and obtains classified application user behavior data, where the classified application user behavior data includes, but is not limited to, lost new user data, user operation data of successful service operation, and old user complex service operation data, for example: taking the service operation as an example for purchase, the classified application behavior data includes operation record data of new lost users (i.e., new lost user data), operation record data before purchase of new users successfully purchased (i.e., user data successfully operated by the service), operation record data after purchase of new users successfully purchased (i.e., user operation data successfully operated by the service), daily record data of new users successfully purchased (i.e., user operation data successfully operated by the service), operation record data before loss of old users (i.e., old user re-service operation data), operation record data before re-purchase of old users (i.e., old user re-service operation data), and operation record data after re-purchase of old users (i.e., old user re-service operation data);

and classifying the classified application user behavior data according to a preset type (the preset type comprises an application operation data type and an application user data type) to obtain initial application operation data and initial application data. The application operation data type is used for indicating operation path data generated by a user operating on an application program, namely application operation data. The application user data type is used to indicate application data generated by a user operating on an application, i.e., application data. The initial application operation data includes, but is not limited to, a user identification number (ID), an operation time, and an operation path. The initial application data includes, but is not limited to, login data, operational application plate data, and tag data.

Specifically, before step 201, the server acquires historical operation sequence data, target user portrait data and recall configuration information of a user based on an application program to be processed, wherein the target user portrait data comprises a score of the operation data; based on historical operation sequence data and recall configuration information, index retrieval and reading are carried out on target user portrait data to obtain a score of the read operation data; scoring the historical operation sequence data according to the score of the read operation data to obtain scored historical operation sequence data; and according to the value of the score, performing descending arrangement on the scored historical operation sequence data to obtain a preset adjustment strategy sequence.

The server acquires operation sequence data of a user based on a historical time period of the application program to be processed in real time after obtaining user authorization, and historical operation sequence data are obtained; extracting pre-created target user portrait data and acquiring recall configuration information corresponding to the target user portrait data, wherein the target user portrait data is used for indicating portrait data of a user corresponding to application user behavior data, the target user portrait data comprises operation data and a score of the operation data, and the recall configuration information comprises the number of recalls and the number of sequential reading; creating an index of historical operation sequence data, recalling the target user portrait data through the index to obtain recall operation data, sequentially reading the recall operation data according to recall configuration information to obtain read operation data, and extracting the score of the read operation data to obtain the score of the read operation data; clustering preset user portrait data based on the read operation data to obtain similar operation data, wherein the preset user portrait data are portrait data of other applicants except users corresponding to the application user behavior data; matching historical operation sequence data through similar operation data and read operation data to obtain target operation data; calculating the sum of the score of the target operation data and the score of the read operation data to obtain a comprehensive score; grading the historical operation sequence data through the comprehensive scores to obtain graded historical operation sequence data; and performing descending arrangement on the scored historical operation sequence data according to the score value to obtain a preset adjustment strategy sequence. The data quality of the preset adjustment strategy sequence is ensured, and the accuracy of the service operation adjustment strategy to be pushed is improved.

202. And storing the initial application operation data based on a linked list in a preset time period to obtain an application operation record sequence.

And the server calls a preset linked list function and stores the initial user application operation data as a linked list according to a preset time period, thereby obtaining a user application operation record sequence. The application operation recording sequence comprises a user identification number, an operation time stamp and operation recording data, and is used for indicating the operation recording data which dynamically changes along with time, namely dynamic time sequence data.

203. And carrying out operation object classification and record statistics based on preset dimensionality on the initial application data to obtain application data to be processed.

The server obtains an operation object of the initial application data, classifies the operation object according to a preset dimension to obtain the classified initial application data, counts the classified initial application data according to a preset statistical index to obtain application data to be processed, and classifies and counts the initial application data with a fixed function to realize classification processing of medium granularity.

The operation object is used for indicating to click a button or a function area corresponding to a plate in the application program, the operation object can be a link, the preset dimensions include but are not limited to the plate, a content carrier, a label and the like of the application program, the preset statistical indexes include but are not limited to the login duration, the login times, the operation times, the residence time of each plate, the preference label and the like of the application program, and the application data to be processed is used for indicating data counted at the same time, namely static data.

The method and the device realize fine analysis granularity and fine classification of the whole user, can perform deep analysis on the operation behavior of a single user, and realize classification of static data and dynamic time sequence data of application user behavior data. The initial application data is subjected to statistical analysis through the procedural indexes, and the prediction accuracy of the business operation intention is improved.

204. And predicting subsequent business operation intention of the application operation record sequence through a preset cyclic neural network model to obtain business operation intention prediction data.

Specifically, the server performs subsequent business operation prediction based on a business operation result on the application operation record sequence through a preset recurrent neural network model to obtain target subsequent business operation information, wherein the target subsequent business operation information comprises target subsequent business operation data and a probability value of the target subsequent business operation data based on the business operation result; acquiring subsequent operation data of the application user behavior data, and matching the subsequent operation data of the application user behavior data with target subsequent service operation data to obtain matched target subsequent service operation data; and determining the probability value corresponding to the matched target subsequent business operation data as business operation intention prediction data.

The method comprises the steps that a server calls a preset recurrent neural network model, sequence feature extraction and activation function operation based on a memory state are carried out on an application operation recording sequence through an input layer, a plurality of hidden layers and an output layer in the recurrent neural network model to obtain target subsequent service operation information, the target subsequent service operation information comprises target subsequent service operation data and a probability value of the target subsequent service operation data based on a service operation result, and the target subsequent service operation data comprises optimal subsequent operation and suboptimal subsequent operation; classifying the service operation based on each current state (subsequent operation) of the application operation record sequence through the target subsequent service operation information, and marking the probability of the subsequent operation to obtain service operation intention prediction data; acquiring subsequent operation data of the application user behavior data, calling a preset matching algorithm, calculating the similarity between the subsequent operation data and target subsequent business operation data, sorting the similarity in a descending order, and determining the target subsequent business operation data corresponding to the sorted first similarity as the matched target subsequent business operation data, wherein the subsequent operation data is used for indicating each subsequent operation based on the service operation achievement in the current state in the application user behavior data, and the subsequent operation is, for example: the current state is that the operation of order pre-payment is performed, and the subsequent operation is payment when the purchase is completed (namely the business operation is achieved); and determining the probability value corresponding to the matched target subsequent business operation data as business operation intention prediction data.

205. Calling a preset dynamic classification model, classifying the application operation record sequence based on business operation intention to obtain dynamic score data, calling a preset static classification model, and classifying the application data to be processed based on the business operation intention to obtain static score data.

Specifically, the server calls a first classification model in preset dynamic classification models, and performs attention feature extraction, probability value calculation and probability value discrimination based on business operation intention on an application operation record sequence to obtain attention score data; performing feature extraction and classification on the application operation record sequence through a second classification model in the dynamic classification model to obtain original score data; summing the attention score data and the original score data to obtain dynamic score data; and calling a preset static classification model, and performing feature extraction, attention feature fusion, probability value calculation and probability value discrimination based on business operation intention on the application data to be processed to obtain static score data.

The preset dynamic classification model comprises a first classification model and a second classification model, the first classification model is a dynamic prediction model based on an attention mechanism, and the second classification model is a dynamic prediction model of a non-attention mechanism. The execution sequence of the dynamic classification model and the static classification model is not limited, and the dynamic classification model can be operated firstly and then the static classification model can be operated, or the dynamic classification model and the static classification model can be operated simultaneously, or the static classification model can be operated firstly and then the dynamic classification model can be operated.

The server calls a first classification model in a preset dynamic classification model, performs feature extraction on an application operation record sequence based on an attention mechanism to obtain attention time sequence features, performs probability value calculation and probability value discrimination based on business operation intention on the attention time sequence features to obtain attention score data, performs feature extraction on the application operation record sequence through a second classification model to obtain initial time sequence features, performs probability value calculation and probability value discrimination based on the business operation intention on the initial time sequence features to obtain original score data, and performs matrix conversion and matrix weighted summation on the attention score data and the original score data to obtain dynamic score data;

the server calls a preset static classification model, performs feature extraction on application data to be processed to obtain initial user application features, performs attention moment array operation on the initial user application features based on an attention mechanism to obtain attention user application features, performs matrix splicing on the initial user application features and the attention user application features to obtain fusion user application features, and performs probability value calculation and probability value discrimination based on business operation intention on the fusion user application features to obtain static score data.

Through the operation, the original information of the original characteristics can be kept, the bias information of the attention characteristics can be kept, high-value information can be quickly screened out from a large amount of characteristic information by using limited attention resources, and the accuracy of the dynamic score data and the static score data is improved.

206. And summing the dynamic score data and the static score data to obtain target score data, and acquiring a service operation adjustment strategy to be pushed from a preset adjustment strategy sequence according to the service operation intention prediction data and the target score data.

Specifically, the server calculates the sum of the dynamic score data and the static score data according to a preset weight to obtain target score data; respectively comparing and analyzing the target score data with business operation intention prediction data and a preset threshold value to obtain intention analysis results and threshold value analysis results; and if the intention analysis result is that the service operation intention prediction data is smaller than the target score data and/or the threshold analysis result is that the target score data is smaller than the preset threshold, performing first adjustment strategy extraction on a preset adjustment strategy sequence to obtain a service operation adjustment strategy to be pushed, wherein the preset adjustment strategy sequence is an adjustment strategy set which performs reverse ordering according to the size of the score of the operation.

The server performs two-dimensional matrix conversion on the dynamic score data and the static score data to obtain a dynamic two-dimensional matrix and a static two-dimensional matrix, and calculates a weighted sum (or weighted arithmetic mean) of the dynamic two-dimensional matrix and the static two-dimensional matrix according to a preset weight to obtain target score data; respectively comparing and analyzing the target score data with business operation intention prediction data and a preset threshold value to obtain intention analysis results and threshold value analysis results; if the intention analysis result is that the service operation intention prediction data is smaller than the target score data and/or the threshold analysis result is that the target score data is smaller than the preset threshold, acquiring a first adjustment strategy from a preset adjustment strategy sequence to obtain a service operation adjustment strategy to be pushed, wherein the preset adjustment strategy sequence is an adjustment strategy set which is sorted according to the sequence of operation score values from large to small; and if the intention analysis result is that the service operation intention prediction data is greater than or equal to the target score data, performing reinforcement learning processing on the recurrent neural network model, and if the threshold analysis result is that the target score data is greater than or equal to a preset threshold, not performing processing.

Specifically, if the intention analysis result is that the business operation intention prediction data is larger than or equal to the target score data, the server performs reinforcement learning on the recurrent neural network model.

If the intention analysis result is that the service operation intention prediction data is larger than or equal to the target score data, the server calls a preset Policy Gradient (PG) deep learning algorithm to adjust the learning rate, update the gradient and distribute the reward of an activation function to the recurrent neural network model, so that the enhanced learning of the recurrent neural network model is realized. Or carrying out forward reinforcement learning or reverse reinforcement learning on the cyclic neural network model to realize reinforcement learning on the cyclic neural network model. The path habits of the users with strong purchasing intentions are further identified by performing reinforcement learning on the recurrent neural network model, so that how to learn to better influence the users with weak purchasing intentions is ensured, the accuracy and comprehensiveness of prediction of business operation intentions are ensured, and the accuracy of business operation adjustment strategies based on application user behavior data analysis is improved.

In the embodiment of the invention, the application user behavior data is analyzed by combining the angles of machine learning and time series deep learning, the accuracy and comprehensiveness of the prediction result of the business operation intention are ensured, the fine analysis granularity and the fine classification of the whole user are realized, the deep analysis can be performed on the operation behavior of a single user, the classification of the static data and the dynamic time series data of the application user behavior data is realized, the prediction accuracy of the business operation intention is improved, and the accuracy of the business operation adjustment strategy based on the analysis of the application user behavior data is further improved.

With reference to fig. 3, the above description is provided for a processing method of application user behavior data in an embodiment of the present invention, and a processing apparatus of application user behavior data in an embodiment of the present invention is described below, where an embodiment of the processing apparatus of application user behavior data in an embodiment of the present invention includes:

the statistical module 301 is configured to obtain application user behavior data, and perform preset type classification, sequence data conversion, and classification statistics on the application user behavior data to obtain an application operation record sequence and application data to be processed;

the prediction module 302 is configured to predict the subsequent business operation intention of the application operation record sequence through a preset recurrent neural network model, so as to obtain business operation intention prediction data;

the classification module 303 is configured to call a preset dynamic classification model, perform classification based on business operation intention on the application operation record sequence to obtain dynamic score data, call a preset static classification model, and perform classification based on business operation intention on application data to be processed to obtain static score data;

the first obtaining module 304 is configured to sum the dynamic score data and the static score data to obtain target score data, and obtain a to-be-pushed business operation adjustment policy from a preset adjustment policy sequence according to the business operation intention prediction data and the target score data.

The function implementation of each module in the processing apparatus for applying user behavior data corresponds to each step in the processing method embodiment for applying user behavior data, and the function and implementation process thereof are not described in detail herein.

Referring to fig. 4, another embodiment of a processing device for applying user behavior data according to an embodiment of the present invention includes:

wherein, the statistic module 301 specifically includes:

the classification unit 3011 is configured to obtain application user behavior data, classify the application user behavior data according to a service operation result and a preset type, and obtain initial application operation data and initial application data, where the preset type includes an application operation data type and an application user data type;

the storage unit 3012 is configured to perform linked list storage based on a preset time period on the initial application operation data to obtain an application operation record sequence;

a statistical unit 3013, configured to perform operation object classification and record statistics based on preset dimensions on the initial application data to obtain application data to be processed;

Optionally, the prediction module 302 may be further specifically configured to:

acquiring subsequent operation data of the application user behavior data, and matching the subsequent operation data of the application user behavior data with target subsequent service operation data to obtain matched target subsequent service operation data;

Optionally, the classification module 303 may be further specifically configured to:

calling a first classification model in a preset dynamic classification model, extracting attention characteristics of an application operation record sequence, calculating probability value based on business operation intention and judging the probability value to obtain attention score data;

Optionally, the first obtaining module 304 includes:

a calculating unit 3041, configured to calculate a sum of the dynamic score data and the static score data according to a preset weight, to obtain target score data;

an analyzing unit 3042, configured to compare and analyze the target score data with the business operation intention prediction data and the preset threshold respectively, to obtain an intention analysis result and a threshold analysis result;

the extracting unit 3043 is configured to, if the intention analysis result is that the service operation intention prediction data is smaller than the target score data, and/or the threshold analysis result is that the target score data is smaller than the preset threshold, perform first adjustment policy extraction on the preset adjustment policy sequence to obtain a service operation adjustment policy to be pushed, where the preset adjustment policy sequence is an adjustment policy set that performs reverse ordering according to the size of the score of the operation.

Optionally, the first obtaining module 304 further includes:

the reinforcement learning unit 3044 is configured to perform reinforcement learning on the recurrent neural network model if the intention analysis result is that the business operation intention prediction data is greater than or equal to the target score data.

Optionally, the processing apparatus for applying user behavior data further includes:

a second obtaining module 305, configured to obtain historical operation sequence data of a user based on the application to be processed, target user portrait data and recall configuration information, where the target user portrait data includes a score of the operation data;

the reading module 306 is used for performing index retrieval and reading on the target user portrait data based on the historical operation sequence data and the recall configuration information to obtain the value of the read operation data;

a scoring module 307, configured to score the historical operation sequence data according to the score of the read operation data, so as to obtain scored historical operation sequence data;

and the sorting module 308 is configured to sort the scored historical operation sequence data in a descending order according to the score value to obtain a preset adjustment strategy sequence.

The function implementation of each module and each unit in the processing apparatus for applying user behavior data corresponds to each step in the embodiment of the processing method for applying user behavior data, and the function and implementation process are not described in detail herein.

Fig. 3 and fig. 4 above describe in detail the processing apparatus for applying user behavior data in the embodiment of the present invention from the perspective of the modular functional entity, and in the following, describe in detail the processing device for applying user behavior data in the embodiment of the present invention from the perspective of hardware processing.

Fig. 5 is a schematic structural diagram of a processing device for applying user behavior data, where the processing device 500 for applying user behavior data may generate relatively large differences due to different configurations or performances, and may include one or more processors (CPUs) 510 (e.g., one or more processors) and a memory 520, and one or more storage media 530 (e.g., one or more mass storage devices) for storing applications 533 or data 532. Memory 520 and storage media 530 may be, among other things, transient or persistent storage. The program stored on the storage medium 530 may include one or more modules (not shown), each of which may include a series of instructions operating on the processing device 500 that apply the user behavior data. Further, the processor 510 may be configured to communicate with the storage medium 530, and execute a series of instruction operations in the storage medium 530 on the processing device 500 applying the user behavior data.

The processing device 500 for applying user behavior data may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input-output interfaces 560, and/or one or more operating systems 531, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, and the like. Those skilled in the art will appreciate that the processing device architecture of the application user behavior data shown in fig. 5 does not constitute a limitation of the processing device of the application user behavior data, and may include more or fewer components than shown, or some components in combination, or a different arrangement of components.

The present application further provides a processing device for applying user behavior data, including: a memory having instructions stored therein and at least one processor, the memory and the at least one processor interconnected by a line; the at least one processor calls the instructions in the memory to enable the processing device applying the user behavior data to execute the steps in the processing method applying the user behavior data. The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium, and which may also be a volatile computer-readable storage medium, having stored therein instructions, which, when executed on a computer, cause the computer to perform the steps of the processing method of application user behavior data.

Further, the computer-readable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to the use of the blockchain node, and the like.

The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.

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

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

The above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims

1. A processing method of application user behavior data is characterized by comprising the following steps:

acquiring application user behavior data, and sequentially performing preset type classification, sequence data conversion and classification statistics on the application user behavior data according to the sequence to obtain an application operation record sequence and application data to be processed;

the acquiring of the application user behavior data, and performing preset type classification, sequence data conversion and classification statistics on the application user behavior data to obtain an application operation recording sequence and application data to be processed includes:

performing operation object classification and record statistics based on preset dimensionality on the initial application data to obtain application data to be processed;

calling a preset dynamic classification model, classifying the application operation record sequence based on business operation intention to obtain dynamic score data, calling a preset static classification model, classifying the application data to be processed based on business operation intention to obtain static score data, wherein the dynamic classification model is used for classifying data with large operation record data amount and time sequence, subsequent operations classify the data with certain dependency relationship on the previous overall operations, and the static classification model is used for classifying the user business execution conditions of the data counted at the same moment;

2. The method for processing application user behavior data according to claim 1, wherein the predicting the subsequent business operation intention of the application operation record sequence through a preset recurrent neural network model to obtain business operation intention prediction data comprises:

3. The method for processing application user behavior data according to claim 1, wherein the step of calling a preset dynamic classification model to classify the application operation record sequence based on business operation intention to obtain dynamic score data and calling a preset static classification model to classify the application data to be processed based on business operation intention to obtain static score data comprises:

calling a first classification model in a preset dynamic classification model, and performing attention feature extraction, probability value calculation and probability value discrimination on the application operation record sequence based on business operation intention to obtain attention score data, wherein the first classification model is a dynamic prediction model based on an attention mechanism;

performing feature extraction and classification on the application operation record sequence through a second classification model in the dynamic classification model to obtain original score data, wherein the second classification model is a dynamic prediction model of a non-attention mechanism;

4. The method for processing user behavior data according to claim 1, wherein the step of summing the dynamic score data and the static score data to obtain target score data, and obtaining a business operation adjustment policy to be pushed from a preset adjustment policy sequence according to the business operation intention prediction data and the target score data comprises:

5. The method for processing application user behavior data according to claim 4, wherein if the intention analysis result indicates that the business operation intention prediction data is smaller than the target score data and/or the threshold analysis result indicates that the target score data is smaller than a preset threshold, performing first adjustment policy extraction on a preset adjustment policy sequence, and after obtaining the business operation adjustment policy to be pushed, further comprising:

6. The method for processing application user behavior data according to any one of claims 1 to 5, wherein before obtaining the application user behavior data, performing preset type classification, sequence data conversion, and classification statistics on the application user behavior data, and obtaining the application operation record sequence and the application data to be processed, the method further comprises:

7. A processing apparatus for applying user behavior data, the processing apparatus for applying user behavior data comprising:

the statistical module is used for acquiring application user behavior data, and sequentially performing preset type classification, sequence data conversion and classification statistics on the application user behavior data according to the sequence to obtain an application operation record sequence and application data to be processed;

the classification module is used for calling a preset dynamic classification model, classifying the application operation record sequence based on business operation intention to obtain dynamic score data, calling a preset static classification model, classifying the application data to be processed based on the business operation intention to obtain static score data, wherein the dynamic classification model is used for classifying data with large operation record data amount and time sequence, subsequent operation has certain dependency relationship on the previous overall operation, and the static classification model is used for classifying the user business execution condition of the data counted at the same time;

8. A processing device for applying user behavior data, the processing device for applying user behavior data comprising: a memory and at least one processor, the memory having instructions stored therein;

the at least one processor invokes the instructions in the memory to cause the processing device applying the user behavior data to execute the processing method applying the user behavior data according to any one of claims 1 to 6.

9. A computer-readable storage medium having instructions stored thereon, wherein the instructions, when executed by a processor, implement a method for processing application user behavior data according to any one of claims 1-6.