User behavior determination method and device, equipment and computer storage medium

Abstract

The invention discloses a user behavior determination method and device, equipment and a computer storage medium, and relates to the technical field of information processing. The method comprises the following steps: acquiring behavior data of a user; inputting the behavior data into a behavior analysis model, performing feature relationship extraction on the behavior feature map through a convolutional neural network, and generating a behavior feature map based on the extracted features; performing graph feature extraction on the behavior feature graph through a graph convolutional network to obtain graph structure features of the behavior feature graph; performing feature graph reconstruction on the graph structure features through an auto-encoder to obtain a reconstructed feature graph; determining a user behavior classification value corresponding to the reconstructed feature map through a full connection layer; and determining the user behavior according to the user behavior classification value. According to the embodiment of the invention, through a hybrid neural network architecture composed of the convolutional neural network, the graph convolutional network and the variational auto-encoder, multi-dimensional features with complex relationships can be extracted and the user behaviors can be determined, so that the accuracy of user behavior analysis is improved.

Description

Technical Field

[0001] The present application belongs to the field of information processing technology, and in particular, relates to a method, device, equipment and computer storage medium for determining user behavior. Background Art

[0002] User behavior analysis occupies an increasingly important position in the field of information processing technology. By analyzing user behavior, related companies can provide personalized services, thereby improving user experience and satisfaction.

[0003] Existing user behavior analysis methods often input user behavior data into a convolutional neural network after obtaining user behavior data, and extract user behavior features through the convolutional neural network. However, since user behavior data is often multi-dimensional and has correlations, the user behavior features extracted by existing methods are often single-dimensional, and cannot extract multi-dimensional features with complex relationships. The user behavior determined by the features extracted by the existing methods is inaccurate, which leads to low accuracy of user behavior analysis. Summary of the invention

[0004] The embodiments of the present application provide a method, apparatus, device and computer storage medium for determining user behavior to solve the problem of low accuracy of user behavior analysis in the prior art.

[0005] In a first aspect, an embodiment of the present application provides a method for determining user behavior, the method comprising:

[0006] Obtain user behavior data;

[0007] The behavior data is input into the behavior analysis model, features of the behavior data are extracted through the convolutional neural network of the behavior analysis model, and a behavior feature graph is generated based on the extracted features;

[0008] The graph convolutional network of the behavior analysis model is used to extract the feature relationships of the features in the behavior feature graph, and the graph structure features of the behavior feature graph are obtained;

[0009] The autoencoder of the behavior analysis model reconstructs the feature map of the graph structure feature to obtain a reconstructed feature map. The reconstructed feature map is a feature map constructed based on the latent vector determined by the autoencoder according to the eigenvalue of the graph structure feature;

[0010] Determine the user behavior classification value corresponding to the reconstructed feature graph through the fully connected layer of the behavior analysis model;

[0011] Determine user behavior based on the user behavior classification value.

[0012] In a second aspect, an embodiment of the present application provides a device for determining user behavior, the device comprising:

[0013] The acquisition module is used to obtain the user's behavior data;

[0014] An extraction module, used to input the behavior data into the behavior analysis model, extract features of the behavior data through the convolutional neural network of the behavior analysis model, and generate a behavior feature map based on the extracted features;

[0015] The extraction module is also used to extract feature relationships from features in the behavior feature graph through the graph convolution network of the behavior analysis model to obtain graph structure features of the behavior feature graph;

[0016] A reconstruction module is used to reconstruct a feature map of the graph structure feature through the autoencoder of the behavior analysis model to obtain a reconstructed feature map. The reconstructed feature map is a feature map constructed based on the latent vector determined by the autoencoder according to the eigenvalue of the graph structure feature;

[0017] A determination module, used to determine the user behavior classification value corresponding to the reconstructed feature graph through the fully connected layer of the behavior analysis model;

[0018] The determination module is also used to determine the user behavior according to the user behavior classification value.

[0019] In a third aspect, an embodiment of the present application provides a terminal device, comprising: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, a method for determining user behavior as in the first aspect is implemented.

[0020] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, a method for determining user behavior as in the first aspect is implemented.

[0021] In a fifth aspect, an embodiment of the present application provides a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes a method for determining user behavior as in the first aspect.

[0022] The embodiment of the present application provides a method, device, equipment and computer storage medium for determining user behavior. After obtaining the user's behavior data, the method inputs the user behavior data into a behavior analysis model, which is composed of a convolutional neural network, a graph convolutional network and an autoencoder. First, a behavior feature graph is obtained by extracting features through a convolutional neural network, and local features of the user behavior data can be extracted. The graph feature extraction of the behavior feature graph is performed through a graph convolutional network to obtain the graph structure features of the behavior feature graph, which can effectively process the feature graph structure data and obtain the mutual relationship and dependency between user behaviors. The feature graph is reconstructed by the autoencoder to obtain a reconstructed feature graph, which can learn the potential representation of the behavior data and obtain deeper features of the user. Finally, the user behavior classification value of the reconstructed feature graph is determined by a fully connected layer, and the corresponding user behavior is determined. By analyzing user behavior through a hybrid neural network architecture composed of a convolutional neural network, a graph convolutional network and an autoencoder, user behavior features can be extracted in depth, features with complex relationships in user behavior data can be obtained, and user behavior can be determined, thereby improving the accuracy of user behavior analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0024] Figure 1 It is a flowchart of a method for determining user behavior provided in an embodiment of the present application;

[0025] Figure 2 It is a flowchart of an implementation method of the behavior data preprocessing process provided in an embodiment of the present application;

[0026] Figure 3 It is a flowchart of an implementation method of the behavior analysis model training process provided in an embodiment of the present application;

[0027] Figure 4 This is a flowchart of an implementation method for determining user behavior provided in an embodiment of the present application;

[0028] Figure 5 It is a flowchart of a device for determining user behavior provided in an embodiment of the present application;

[0029] Figure 6 It is a structural diagram of the terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0031] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0032] User behavior analysis is becoming increasingly important in the field of information processing technology because user behavior data contains huge use value. By deeply analyzing user behavior patterns, companies can better understand user needs and provide more personalized services and products to improve user experience and satisfaction. Existing methods can often only extract features of a single dimension, for example, predicting user preferences by analyzing user click frequency or purchase history. This method ignores the multi-dimensional characteristics of user behavior, such as time, location, device type and other dimensions, and there are complex correlations between these dimensions. For example, a user's purchasing behavior may be affected by time, such as holidays, and location, such as residential area, which work together in the user's decision-making process. However, the user behavior features extracted by existing methods are often single-dimensional, and cannot extract multi-dimensional features with complex relationships, making the user behavior determined by the features extracted by existing methods inaccurate, which in turn leads to low accuracy of user behavior analysis.

[0033] In order to solve the existing technology, the embodiment of the present application provides a method, device, equipment and computer storage medium for determining user behavior. After the method obtains the user's behavior data, the user behavior data is input into a behavior analysis model, and the behavior analysis model is composed of a convolutional neural network, a graph convolutional network and an autoencoder. First, the behavior feature graph is obtained by feature extraction through a convolutional neural network, and the local features of the user behavior data can be extracted. The graph feature extraction of the behavior feature graph through a graph convolutional network is performed to obtain the graph structure features of the behavior feature graph, which can effectively process the feature graph structure data and obtain the mutual relationship and dependency between user behaviors. The feature graph is reconstructed by the autoencoder to obtain a reconstructed feature graph, which can learn the potential representation of the behavior data and obtain the deeper features of the user. Finally, the user behavior classification value of the reconstructed feature graph is determined by a fully connected layer, and the corresponding user behavior is determined. By analyzing the user behavior through a hybrid neural network architecture composed of a convolutional neural network, a graph convolutional network and an autoencoder, the user behavior features can be extracted from multiple dimensions and in depth, and the user behavior can be determined by obtaining the complex patterns and potential relationships of the user behavior, thereby improving the accuracy of the user behavior analysis.

[0034] The following first introduces the method for determining user behavior provided in an embodiment of the present application.

[0035] Figure 1 FIG. 1 is a flow chart showing a method for determining user behavior provided by an embodiment of the present application. Figure 1 As shown, the method may include the following steps: S101 to S106.

[0036] S101, obtaining user behavior data.

[0037] Among them, user behavior data refers to the behavior and activity data generated when users interact with products or services online or offline, which can reflect users' habits, preferences, needs and interaction patterns.

[0038] In some embodiments, the user's behavior data may include user call records, text message usage records, data traffic usage records, and recharge records.

[0039] In some embodiments, the user's behavior data may be obtained by using an application programming interface (API), log files, and database exports.

[0040] In one example, an API connection can be established with a user management system of a communication operator and an API request can be sent to obtain user call records from the user management system of the communication operator through the API interface. The user call records may include the call start time, call end time, call duration, and the caller number.

[0041] By configuring the log file parser to parse the log files generated by the SMS gateway system, the user's SMS usage records can be obtained from the communication operator's SMS gateway system. The SMS usage records can include the SMS sending time, SMS recipient number, and SMS length.

[0042] The user's data traffic usage record is obtained from the communication operator's data traffic management system through database export, a connection with the data traffic management system database is established, and a structured query language (SQL) query statement is written to export the user's data traffic usage record from the database. The data traffic usage record may include the start time of data traffic usage, the end time of data traffic usage, the total traffic used, and the type of application used.

[0043] By establishing an API connection with the recharge management system of the communication operator and sending an API request, the user's recharge record is retrieved from the recharge management system of the communication operator through the API interface. The user's recharge record may include the recharge time, recharge amount and recharge method.

[0044] S102, inputting the behavior data into the behavior analysis model, extracting features of the behavior data through the convolutional neural network of the behavior analysis model, and generating a behavior feature graph based on the extracted features.

[0045] Among them, the convolutional neural network includes a convolution layer, a pooling layer, and a fully connected layer. On the convolution layer, a pre-trained convolution kernel is used to perform a convolution operation on the behavior data, extract the features of the behavior data, and generate a feature map; the feature map is pooled through the pooling layer to reduce the spatial size of the feature map. The convolutional neural network can include multiple convolutional layers and pooling layers. After being processed by multiple convolutional layers and pooling layers, the feature map is input into the fully connected layer to perform weighted summation of the features of the feature map according to the preset weights to obtain a behavior feature map.

[0046] Convolutional neural networks extract features from input data to generate behavioral feature maps, which can automatically identify features in behavioral data and convert raw behavioral data into high-level abstract feature representations.

[0047] S103, extracting feature relationships of features in the behavior feature graph through the graph convolutional network of the behavior analysis model to obtain graph structure features of the behavior feature graph.

[0048] Among them, the graph convolution network is composed of a graph convolution layer, a pooling layer, and an output layer. The graph convolution network can include multiple graph convolution layers and pooling layers. The graph convolution layer is used to propagate information and aggregate features on the graph structure. Each graph convolution layer receives the node feature matrix and adjacency matrix corresponding to the graph structure as input, and updates the feature representation of the node through convolution operations; the pooling layer is used to aggregate node features to reduce feature dimensions; the output layer is used to generate the final output of the model.

[0049] In some embodiments, extracting graph features from the behavior feature graph through a graph convolutional network of a behavior analysis model to obtain graph structural features of the behavior feature graph may include:

[0050] Construct a graph structure based on the behavior feature graph. The nodes in the graph structure are feature vectors in the behavior feature graph, and the edges in the graph structure are the similarities or spatial relationships between the features corresponding to the nodes.

[0051] Determine the adjacency matrix and feature matrix based on the graph structure. Each element of the adjacency matrix indicates whether there is an edge between nodes, and each element of the feature matrix indicates the feature vector of a node.

[0052] Multiply the adjacency matrix and the feature matrix to obtain the adjacent node feature matrix. Each element of the adjacent node matrix represents the adjacent node feature of each node.

[0053] The adjacent node feature matrix is ​​transformed according to the pre-trained weight matrix and activation function, and the node features are updated using the transformed feature matrix to obtain the graph structure features.

[0054] By using a graph convolutional network, the adjacency matrix and feature matrix of the graph structure are multiplied to obtain the weighted sum of the adjacent node features of each node, and the adjacent node feature information of each node is aggregated to the node itself, so that the node can capture the relationship between each node and the adjacent node; using pre-trained weight matrices and activation functions can extract more effective feature information and learn deeper feature representations in the data.

[0055] S104, reconstructing a feature map of the graph structure feature through the autoencoder of the behavior analysis model to obtain a reconstructed feature map, where the reconstructed feature map is a feature map constructed based on the latent vector determined by the autoencoder according to the eigenvalues ​​of the graph structure feature.

[0056] Among them, the autoencoder is an unsupervised learning neural network, including an encoder and a decoder. The encoder encodes the input data into a low-dimensional representation, namely a latent vector, and then the decoder reconstructs the original data from the latent vector.

[0057] Since graph structure features can represent the complex relationships of behavioral data, the reconstructed feature graph obtained by reconstructing the graph structure features can reflect the complex relationships of behavioral data; the encoder reduces the high-dimensional features that reflect the complex relationships of behavioral data to low-dimensional latent vectors through the encoding process, which can remove redundant and noise information and retain key features; the latent vector is then reconstructed into a feature graph through the decoder, so that the features in the reconstructed feature graph are the key features for removing redundant information, thereby obtaining deeper features.

[0058] S105, determining the user behavior classification value corresponding to the reconstructed feature graph through the fully connected layer of the behavior analysis model.

[0059] Among them, the user behavior classification value represents the possibility of the user being classified into the corresponding user behavior category, which can be a classification probability, and each user behavior classification value corresponds to a user behavior classification category. The fully connected layer has multiple neurons, and the corresponding user behavior classification values ​​of multiple classification categories can be determined through the fully connected layer. Each neuron in the fully connected layer corresponds to a user behavior classification value of a user behavior category, and each neuron performs weighted summation of the input through pre-trained weights to obtain the corresponding user behavior classification value.

[0060] By performing weighted summation on the input through the preset weights of the fully connected layer, the features in the reconstructed feature map can be integrated and mapped to the user behavior classification value, thus realizing the mapping from feature space to category space.

[0061] S106, determining the user behavior according to the user behavior classification value.

[0062] In some embodiments, the user behavior may be determined based on a preset threshold, and when the user behavior classification value exceeds the preset threshold of the corresponding classification category, the user is judged to have the behavior of the classification category.

[0063] In some embodiments, the classification category corresponding to the maximum user behavior classification value may be determined as the user behavior.

[0064] By extracting features through convolutional neural networks to obtain behavior feature graphs, local features of user behavior data can be extracted. By extracting graph features from behavior feature graphs through graph convolutional networks, graph structure features of behavior feature graphs can be obtained, which can effectively process feature graph structure data and obtain the relationships and dependencies between user behaviors. By reconstructing the feature graph of graph structure features through autoencoders to obtain reconstructed feature graphs, potential representations of behavior data can be learned to obtain deeper features of users. Finally, the user behavior classification value of the reconstructed feature graph is determined through the fully connected layer, and the corresponding user behavior is determined. By analyzing user behavior through a hybrid neural network architecture consisting of convolutional neural networks, graph convolutional networks, and autoencoders, multiple dimensional features extracted from convolutional neural networks can be associated to obtain deep user behavior features that can reflect feature associations, and user behavior can be determined based on the user behavior features, thereby improving the accuracy of user behavior analysis.

[0065] In some embodiments, when the convolutional neural network extracts features from the behavior data, it outputs a behavior feature graph F cnn The formula can be:

[0066]

[0067] Among them, W i is the convolution kernel, x i is the input data, b i is the bias term, and * is the convolution operation.

[0068] In some embodiments, the graph feature extraction is performed on the behavior feature graph, and the formula for obtaining the graph structure feature of the behavior feature graph may be:

[0069] F gcn =σ(D -1 / 2 AD -1 / 2 F cnn W gcn )

[0070] Where A is the adjacency matrix of the graph. If there is an edge between node i and node j, then A ij The value is 1, otherwise it is 0, D is the degree matrix, W gcn is the weight matrix of the graph convolution layer, and σ is the activation function.

[0071] In some embodiments, reconstructing a feature graph of the graph structure features by an autoencoder to obtain a behavior feature graph may include:

[0072] Calculate the mean and standard deviation of graph structural features;

[0073] Determine the latent vector based on the mean and standard deviation of the graph structure features;

[0074] The feature map is reconstructed according to the latent vector to obtain a reconstructed feature map.

[0075] In some embodiments, the autoencoder used may be a variational autoencoder.

[0076] In some embodiments, the latent vector is determined according to the mean and standard deviation of the graph structure features, and the formula may be:

[0077] z=μ+σ·∈

[0078] μ=f μ (F gcn )

[0079] σ=f σ (F gcn )

[0080] Among them, z is the latent vector, μ is the mean, and f μ is the mean function, σ is the standard deviation, f σ is the standard deviation function, ∈ is the noise sampled from the standard normal distribution.

[0081] In some embodiments, determining the user behavior classification value corresponding to the reconstructed feature graph through the fully connected layer of the behavior analysis model may include:

[0082] Input the reconstructed feature map into the fully connected layer to obtain the output value of each neuron in the fully connected layer;

[0083] The user behavior classification value of each classification category is determined based on the output value.

[0084] The output value calculation formula of each neuron can be:

[0085] F fc =W fc ·F vae +b fc

[0086] Among them, F vae To reconstruct the feature map, W fc is the weight matrix of the fully connected layer, b fc is the bias term.

[0087] The calculation formula for the user behavior classification value of each classification category can be:

[0088]

[0089] Among them, P i is the classification probability of the i-th category, F fc,i is the output value of the i-th neuron in the fully connected layer, and N is the total number of classification categories.

[0090] In some embodiments, before the behavioral data is input into the behavioral analysis model, Figure 2 As shown, the method may further include: S201 to S206.

[0091] S201, removing target data that meets preset rules from the behavior data.

[0092] Among them, the behavioral data is in the format of a data set, and the target data that conforms to the preset rules may include missing values, unreasonable timestamps, and abnormal numerical range data.

[0093] S202, generating a data identifier of the behavior data with the target data removed according to a preset user identifier and a timestamp of the behavior data.

[0094] In some embodiments, when a data identifier is generated, it can be generated by concatenating a preset user identifier and a timestamp, or it can be generated using a preset user identifier and a timestamp according to a preset conversion rule.

[0095] S203: When the same data identifier exists, remove the repeated behavior data corresponding to the same data identifier to obtain the remaining behavior data.

[0096] S204, converting the remaining behavior data according to a preset format to obtain the remaining behavior data in the preset format.

[0097] In some embodiments, the format conversion includes time format conversion and numerical unit conversion. The time format conversion is to unify the time format of all data records, which can be unified into the ISO 8601 standard format; the numerical unit conversion can be to unify the data flow unit and the call duration unit, and the data flow unit can be unified into MB, and the call duration unit can be unified into seconds.

[0098] S205, align the remaining behavior data in the preset format according to the timestamps to obtain target behavior data.

[0099] The timestamps in the standardized data set are aligned to ensure the consistency of each data record at the same time point.

[0100] S206, inputting the target behavior data into the behavior analysis model, including:

[0101] The target behavior data is used as the behavior data of the input behavior analysis model and is input into the behavior analysis model.

[0102] By preprocessing the data, abnormal and duplicate data are removed to avoid the impact of outliers on model performance; and the data format is unified to ensure data consistency.

[0103] In some embodiments, the target behavior data is in a data set format, and the target behavior data obtained after data preprocessing is stored in a distributed storage system, which uses a Hadoop Distributed File System (HDFS) for data storage and management.

[0104] Storing target behavior data in a distributed storage system may include:

[0105] Configure the distributed file system, create file directories for storing preprocessed data, and configure system parameters to optimize data storage and management;

[0106] According to the configuration of system parameters, the target behavior data set is divided into data blocks of preset size. The size of each data block is 64MB or 128MB. A unique identifier is generated for each data block, making the data block unique and traceable in the distributed storage system.

[0107] Use a distributed data transmission protocol to transfer data blocks from local storage to a distributed storage system, use multi-threaded parallel transmission to improve data import efficiency, and write the data blocks transferred to the distributed storage system into a specified file directory;

[0108] Create multiple copies of each data block in the distributed storage system, distributed on different nodes, and use the distributed storage system monitoring tool to monitor the stored data blocks in real time to detect and repair possible data block damage or loss;

[0109] Configure access permissions for different users and applications, optimize data retrieval strategies, and use the MapReduce big data processing framework to retrieve and process data stored in distributed storage systems.

[0110] Using a distributed storage system for storage, the distributed storage of data blocks on different nodes helps to balance the load and improve resource utilization; and by creating multiple copies of data blocks, the system can use the copies of the remaining nodes when some nodes fail, without causing data loss; configuring access permissions for different users and applications, different users and applications can only access the nodes they have permissions for, enhancing data security and privacy protection.

[0111] In some embodiments, before inputting the behavior data into the behavior analysis model, the method may further include:

[0112] Use the pre-trained feature extraction model to extract features from the behavior data to obtain behavior features;

[0113] Input behavioral data into the behavioral analysis model, including:

[0114] The behavioral characteristics are input into the behavior analysis model.

[0115] Among them, the extracted features may include consumption frequency, usage period, recharge amount, data traffic usage, call duration and number of text messages, etc.

[0116] By extracting features from the data and inputting the extracted features into the behavior analysis model, the behavior analysis model can use the extracted features for analysis and learn deeper and more abstract feature representations from the existing features. Therefore, when the behavior analysis model obtains complex relationship features between data based on deep feature representations, it can improve the accuracy of complex relationship features, thereby improving the accuracy of user behavior analysis.

[0117] In some embodiments, before the behavioral data is input into the behavioral analysis model, Figure 3 As shown, the method may further include: S301 to S304.

[0118] S301, obtaining user behavior training data;

[0119] S302, inputting the behavior training data into a preset analysis model, and determining a behavior training classification value of the behavior training data according to model parameters of the preset analysis model;

[0120] S303, determining a loss function value according to the behavior training classification value and the preset classification value of the behavior training data;

[0121] S304, when the loss function value is greater than the set threshold, redetermine the model parameters according to the loss function value and update the preset analysis model until the loss function value is less than the set threshold, and obtain the preset model parameters and the behavior analysis model corresponding to the preset model parameters.

[0122] By obtaining behavioral training data and adjusting model parameters, the model can be optimized to better fit the actual data and improve the accuracy of predictions.

[0123] In some embodiments, when the model parameters are redetermined according to the loss function value and the preset analysis model is updated, the back propagation algorithm is used to redetermine the model parameters.

[0124] In some embodiments, user behavior is determined based on the user behavior classification value, such as Figure 4 As shown, it may include: S401 to S403.

[0125] S401, obtaining an actual behavior classification value of the user's behavior data;

[0126] S402, determining target model parameters of the behavior analysis model according to the user behavior classification value, the actual behavior classification value and the preset model parameters of the behavior analysis model;

[0127] S403: Update the behavior analysis model according to the target model parameters.

[0128] The model parameters are updated in real time according to the actual behavior classification values, allowing the model to continuously learn and update based on new data, maintaining the timeliness and accuracy of the model.

[0129] In some embodiments, determining the target model parameters of the behavior analysis model according to the user behavior classification value, the actual behavior classification value, and the preset model parameters of the behavior analysis model may include:

[0130] The loss function value is determined based on the user behavior classification value and the actual behavior classification value. The formula can be:

[0131]

[0132] Among them, L recon represents the reconstruction loss, L KL represents the Kullback-Leibler divergence loss, N represents the number of samples, and y i represents the actual behavior classification value, P i represents the user behavior classification value, M represents the dimension of the latent vector, μ j and σ j represents the mean and standard deviation of the latent vector z.

[0133] The target model parameters are determined according to the preset model parameters and the loss function value. The formula can be:

[0134]

[0135] Among them, θ represents the preset model parameters, η represents the learning rate, Denotes the loss function L recon +L KL The gradient with respect to the parameter θ is represents the partial derivative of the loss function L with respect to the parameter θ, q is the number of periodic terms, and ρ l , σ l ,φ l represents the preset adjustment parameters, which are the amplitude, frequency and phase of the lth periodic item, and t represents the timestamp.

[0136] In some embodiments, after determining the user behavior classification value corresponding to the reconstructed feature graph through the fully connected layer of the behavior analysis model, the method further includes:

[0137] The target user demand value corresponding to the user behavior classification value and the timestamp of the user behavior data is determined according to the relationship between the behavior classification value, the timestamp and the user demand value.

[0138] By analyzing the relationship between behavior classification values, timestamps, and user demand values, we can predict users’ future demand for corresponding needs. At the same time, considering timestamp information enables the model to capture the temporal dynamics of user behavior and improve the timeliness of demand forecasting.

[0139] In some embodiments, the relationship between the behavior classification value, the timestamp, and the user demand value may be:

[0140]

[0141] Among them, P(t) represents the demand forecast classification value at time t, φ0 represents the preset constant term, which represents the mean or intercept of the time series, and φ i Represents the autoregressive coefficient, which represents the weight of the i-th lag term in the time series, R real (ti) represents the behavior classification value at time ti, θ j represents the moving average coefficient, represents the weight of the j-th lagged error term in the time series, ∈(t) represents the error term at time t, ∈(tj) represents the error term at time tj, p represents the order of the autoregressive term, and q represents the order of the moving average term.

[0142] In some embodiments, after determining the user behavior classification value corresponding to the reconstructed feature graph, the method may further include:

[0143] Classify and process according to the user behavior classification value to generate a user behavior report R report , can be:

[0144]

[0145] Among them, κ represents the result processing function, which can be a scaling factor, p is the total number of input features, and h k Represents the preset function of the kth feature, R real,k represents the user behavior classification value corresponding to the k-th classification category, that is, the k-th input feature, ν k , ξ, ω, τ are preset model parameters, which respectively represent the weight, contribution parameter, time coefficient and constant term of the kth feature. Among them, the contribution parameter represents the contribution of the exponential term to the entire expression, the time coefficient represents the rate at which the exponential term changes over time, and the constant term represents the initial value or offset of the exponential term.

[0146] The result processing function and the preset function can be set according to user needs. For example, the result processing function can be a threshold judgment function, which determines the user behavior based on whether the sum or weighted sum of the classification values ​​exceeds a certain threshold; it can also be a weighted average function to capture time trends or periodic change results.

[0147] In some embodiments, after determining the user behavior classification value corresponding to the reconstructed feature graph, the method may further include:

[0148] The behavior classification values ​​are processed through the cluster analysis algorithm to identify the user's consumption habits and generate the cluster center C:

[0149]

[0150] Among them, c k represents the kth cluster center, C k represents the kth cluster, and r represents the cluster belonging to C k The sample point of |C k | represents the number of sample points in the kth cluster, and K represents the total number of clusters;

[0151] Use association rule mining technology to process the behavior classification values, identify the user's consumption preferences, and generate association rule set A:

[0152]

[0153] in, Representation Rules , sup(X∪Y) represents the support of X∪Y, α represents the confidence threshold, and β represents the support threshold.

[0154] Figure 5 FIG. 5 is a schematic diagram showing a structure of a device 500 for determining user behavior provided in an embodiment of the present application. Figure 5 As shown, the device may include:

[0155] Acquisition module 501, used to acquire user behavior data;

[0156] An extraction module 502, for inputting the behavior data into the behavior analysis model, extracting features from the behavior data through a convolutional neural network of the behavior analysis model, and generating a behavior feature graph based on the extracted features;

[0157] The extraction module 502 is further used to extract graph features from the behavior feature graph through the graph convolution network of the behavior analysis model to obtain graph structural features of the behavior feature graph;

[0158] The reconstruction module 503 is used to reconstruct the feature map of the graph structure feature through the autoencoder of the behavior analysis model to obtain a reconstructed feature map. The reconstructed feature map is a feature map constructed based on the latent vector determined by the autoencoder according to the eigenvalue of the graph structure feature.

[0159] A determination module 504, configured to determine a user behavior classification value corresponding to the reconstructed feature graph through a fully connected layer of the behavior analysis model;

[0160] The determination module 504 is further configured to determine the user behavior according to the user behavior classification value.

[0161] In some embodiments, the user behavior determination device 500 may further include:

[0162] A removal module is used to remove target data that meets preset rules from the behavior data;

[0163] A generating module, used for generating a data identifier of the behavior data with the target data removed according to a preset user identifier and a timestamp of the behavior data;

[0164] The removal module is further used to remove duplicate behavior data corresponding to the same data identifier when there is an identical data identifier, so as to obtain the remaining behavior data;

[0165] A conversion module, used for converting the remaining behavior data into a format according to a preset format to obtain the remaining behavior data in the preset format;

[0166] An alignment module, used to align the remaining behavior data in a preset format according to timestamps to obtain target behavior data;

[0167] The input module is used to input the target behavior data into the behavior analysis model as the behavior data of the input behavior analysis model.

[0168] In some embodiments, the extraction module 502 is further used to extract features from the behavior data using a pre-trained feature extraction model to obtain behavior features;

[0169] The input module is also used to input behavioral features into the behavior analysis model.

[0170] In some embodiments, the user behavior determination device 500 may further include:

[0171] The acquisition module 501 is also used to acquire the user's behavior training data;

[0172] The determination module 504 is further used to input the behavior training data into a preset analysis model, and determine the behavior training classification value of the behavior training data according to the model parameters of the preset analysis model;

[0173] The determination module 504 is further used to determine the loss function value according to the behavior training classification value and the preset classification value of the behavior training data;

[0174] The updating module is used to redetermine the model parameters and update the preset analysis model according to the loss function value when the loss function value is greater than the set threshold, until the loss function value is less than the set threshold, thereby obtaining the preset model parameters and the behavior analysis model corresponding to the preset model parameters.

[0175] In some embodiments, the acquisition module 501 is further used to obtain the actual behavior classification value of the user's behavior data;

[0176] The determination module 504 is further used to determine the target model parameters of the behavior analysis model according to the user behavior classification value, the actual behavior classification value and the preset model parameters of the behavior analysis model;

[0177] The updating module is also used to update the behavior analysis model according to the target model parameters.

[0178] In some embodiments, the determination module 504 is further used to determine the target user demand value corresponding to the user behavior classification value and the timestamp of the user behavior data according to the relationship between the behavior classification value, the timestamp and the user demand value.

[0179] Figure 5 The various modules in the device shown can be implemented Figure 1 Each step in the above steps can achieve the corresponding technical effect, which will not be repeated here for the sake of brevity.

[0180] Figure 6 A schematic diagram of the hardware structure of a terminal device provided in an embodiment of the present application is shown.

[0181] The terminal device may include a processor 601 and a memory 602 storing computer program instructions.

[0182] Specifically, the processor 601 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0183] The memory 602 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 602 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In one example, the memory 602 may include a removable or non-removable (or fixed) medium, or the memory 602 is a non-volatile solid-state memory. The memory 602 may be inside or outside the integrated gateway disaster recovery device.

[0184] In one example, the memory 602 may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Thus, typically, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method for determining user behavior according to the present disclosure.

[0185] The processor 601 reads and executes the computer program instructions stored in the memory 602 to implement Figure 1 The method of determining user behavior in the illustrated embodiment.

[0186] In one example, the terminal device may further include a communication interface 603 and a bus 604. Figure 6 As shown, the processor 601, the memory 602, and the communication interface 603 are connected via a bus 604 and communicate with each other.

[0187] The communication interface 603 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0188] Bus 604 includes hardware, software or both, and the components of terminal equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (Accelerated Graphics Port, AGP) or other graphics bus, enhanced industry standard architecture (Extended Industry Standard Architecture, EISA) bus, front side bus (Front Side Bus, FSB), Hyper Transport (Hyper Transport, HT) interconnection, industry standard architecture (Industry Standard Architecture, ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 604 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0189] In addition, in combination with the method for determining user behavior in the above embodiments, the present application embodiment may provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by a processor, any one of the methods for determining user behavior in the above embodiments is implemented.

[0190] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed, implements any one of the user behavior determination methods in the above embodiments.

[0191] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0192] The functional blocks shown in the structural block diagram described above can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), appropriate firmware, plug-in, function card, etc. When implemented in software, the elements of the present application are programs or text segments used to perform the required tasks. The program or text segment can be stored in a machine-readable medium, or transmitted on a transmission medium or communication link via a data signal carried in a carrier. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, read-only memory (Read-Only Memory, ROM), flash memory, erasable read-only memory (Erasable ReadOnly Memory, EROM), floppy disks, compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), optical discs, hard disks, optical fiber media, radio frequency (Radio Frequency, RF) links, etc. The text segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0193] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0194] Aspects of the present disclosure are described above with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0195] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A method for determining user behavior, characterized in that: include: Obtain user behavior data; Inputting the behavior data into a behavior analysis model, extracting features from the behavior data through a convolutional neural network of the behavior analysis model, and generating a behavior feature graph based on the extracted features; Extracting feature relationships from features in the behavior feature graph through the graph convolutional network of the behavior analysis model to obtain graph structure features of the behavior feature graph; Reconstructing a feature map of the graph structure feature by the autoencoder of the behavior analysis model to obtain a reconstructed feature map, wherein the reconstructed feature map is a feature map constructed based on the latent vector determined by the autoencoder according to the eigenvalue of the graph structure feature; Determining the user behavior classification value corresponding to the reconstructed feature graph through the fully connected layer of the behavior analysis model; The user behavior is determined according to the user behavior classification value.

2. The method for determining user behavior according to claim 1, characterized in that: Before inputting the behavior data into the behavior analysis model, the method further comprises: Removing target data that meets preset rules from the behavior data; Generate a data identifier of the behavior data from which the target data is removed according to a preset user identifier and a timestamp of the behavior data; In the case where the same data identifier exists, removing the repeated behavior data corresponding to the same data identifier to obtain the remaining behavior data; Converting the remaining behavior data into a preset format to obtain the remaining behavior data in the preset format; Align the remaining behavior data in the preset format according to the timestamp to obtain target behavior data; The step of inputting the behavior data into a behavior analysis model comprises: The target behavior data is used as the behavior data of the input behavior analysis model and is input into the behavior analysis model.

3. The method for determining user behavior according to claim 1, characterized in that: Before inputting the behavior data into the behavior analysis model, the method further comprises: Using a pre-trained feature extraction model to extract features from the behavior data to obtain behavior features; The step of inputting the behavior data into a behavior analysis model comprises: The behavioral characteristics are input into a behavior analysis model.

4. The method for determining user behavior according to claim 1, characterized in that: Before inputting the behavior data into the behavior analysis model, the method further comprises: Obtain user behavior training data; Inputting the behavior training data into a preset analysis model, and determining a behavior training classification value of the behavior training data according to model parameters of the preset analysis model; Determining a loss function value according to the behavior training classification value and a preset classification value of the behavior training data; When the loss function value is greater than a set threshold, the model parameters are re-determined according to the loss function value and the preset analysis model is updated until the loss function value is less than the set threshold, thereby obtaining the preset model parameters and the behavior analysis model corresponding to the preset model parameters.

5. The method for determining user behavior according to claim 1, characterized in that: After determining the user behavior according to the user behavior classification value, the method further includes: Obtain the actual behavior classification value of the user's behavior data; Determining target model parameters of the behavior analysis model according to the user behavior classification value, the actual behavior classification value and the preset model parameters of the behavior analysis model; The behavior analysis model is updated according to the target model parameters.

6. The method for determining user behavior according to claim 1, characterized in that: After determining the user behavior classification value corresponding to the reconstructed feature graph through the fully connected layer of the behavior analysis model, the method further includes: The target user demand value corresponding to the user behavior classification value and the timestamp of the user behavior data is determined according to the relationship between the behavior classification value, the timestamp and the user demand value.

7. A device for determining user behavior, characterized in that: The device comprises: The acquisition module is used to obtain the user's behavior data; An extraction module, used to input the behavior data into a behavior analysis model, perform feature extraction on the behavior data through a convolutional neural network of the behavior analysis model, and generate a behavior feature graph based on the extracted features; The extraction module is further used to extract feature relationships of features in the behavior feature graph through the graph convolution network of the behavior analysis model to obtain graph structure features of the behavior feature graph; A reconstruction module, used to reconstruct a feature map of the graph structure feature through the autoencoder of the behavior analysis model to obtain a reconstructed feature map, wherein the reconstructed feature map is a feature map constructed based on the latent vector determined by the autoencoder according to the eigenvalue of the graph structure feature; A determination module, used to determine the user behavior classification value corresponding to the reconstructed feature graph through the fully connected layer of the behavior analysis model; The determination module is further used to determine the user behavior according to the user behavior classification value.

8. A terminal device, characterized in that: The device comprises: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the method for determining user behavior as described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the method for determining user behavior according to any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the method for determining user behavior as described in any one of claims 1 to 6.

Images