Adaptive spatio-temporal network behavior pattern evolution analysis method and device and storage medium

Abstract

The invention discloses a self-adaptive spatio-temporal network behavior pattern evolution analysis method and device and a storage medium, and the method comprises the steps: collecting agent multi-dimensional behavior data through machine learning, the agent multi-dimensional behavior data comprising financial data, social media data and news report data; performing data quality monitoring according to a user analysis target, and updating the multi-dimensional behavior data of the intelligent agent; performing tensor decomposition on the updated agent multi-dimensional behavior data to obtain target behavior data; according to the target behavior data, a self-adaptive space-time network model is constructed, and the type of the self-adaptive space-time network model comprises a directed graph, an undirected graph or a weighted graph; according to the self-adaptive space-time network model, behavior pattern evolution analysis is carried out, and a behavior pattern evolution analysis result is obtained. The behavior pattern evolution analysis is realized, and the analysis accuracy and the analysis depth are improved. The method can be widely applied to the technical field of behavior data analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of behavioral data analysis, and in particular, to an adaptive spatio-temporal network behavior pattern evolution analysis method, device, and storage medium. Background Art

[0002] An agent refers to an entity participating in transactions or interactions in the financial market, including individual investors, institutional investors, algorithmic trading programs, etc. Analyzing the evolution of the behavior patterns of agents in the financial market is beneficial to understanding the current market trends for making different decisions. Traditional methods use time series analysis or clustering analysis to capture the behavior patterns of agents by analyzing historical data. However, the behavior patterns of agents in the financial market are usually dynamically changing, resulting in low analysis accuracy. At the same time, traditional methods usually only analyze a single agent, but there is a complex relationship network among agents, resulting in low analysis depth.

[0003] In summary, the technical problems existing in the related art need to be improved. Summary of the Invention

[0004] Embodiments of the present invention provide an adaptive spatio-temporal network behavior pattern evolution analysis method, device, and storage medium, which effectively improve the analysis accuracy and depth.

[0005] On the one hand, embodiments of the present invention provide an adaptive spatio-temporal network behavior pattern evolution analysis method, including the following steps:

[0006] Collect multi-dimensional behavioral data of agents using machine learning, where the multi-dimensional behavioral data of agents includes financial data, social media data, and news report data;

[0007] According to the user's analysis target, perform data quality monitoring and update the multi-dimensional behavioral data of agents;

[0008] Perform tensor decomposition on the updated multi-dimensional behavioral data of agents to obtain target behavioral data;

[0009] Construct an adaptive spatio-temporal network model according to the target behavioral data, where the type of the adaptive spatio-temporal network model includes a directed graph, an undirected graph, or a weighted graph;

[0010] Perform behavior pattern evolution analysis according to the adaptive spatio-temporal network model to obtain a behavior pattern evolution analysis result.

[0011] In some embodiments, the collecting multi-dimensional behavioral data of agents using machine learning includes:

[0012] Identify the importance degree of data sources using machine learning;

[0013] Set the data collection frequency and data collection range according to the importance level of the data source;

[0014] Collect the multi-dimensional behavior data of the agent according to the data collection frequency and the data collection range.

[0015] In some embodiments, the performing data quality monitoring according to the user analysis objective and updating the multi-dimensional behavior data of the agent includes:

[0016] Perform data quality analysis on the multi-dimensional behavior data of the agent according to the user analysis objective to obtain a data quality analysis result, where the data quality analysis includes integrity analysis, accuracy analysis, consistency analysis, and timeliness analysis;

[0017] If the data quality analysis result is that the quality does not meet the standard, re-collect data from the data backup source and update the multi-dimensional behavior data of the agent.

[0018] In some embodiments, the performing tensor decomposition on the updated multi-dimensional behavior data of the agent to obtain target behavior data includes:

[0019] Extract features from the updated multi-dimensional behavior data of the agent to obtain a core tensor;

[0020] Determine the data dimension according to the updated multi-dimensional behavior data of the agent;

[0021] Set a factor matrix corresponding to each dimension according to the data dimension;

[0022] Generate a reconstructed tensor as the target behavior data according to the core tensor and the multiple factor matrices.

[0023] In some embodiments, the constructing an adaptive spatio-temporal network model according to the target behavior data includes:

[0024] Set network nodes according to the target behavior data, where the network nodes include agents or agent behaviors;

[0025] Set network edges according to the target behavior data and the network nodes, where the network edges are used to represent the associations between the network nodes;

[0026] Set an edge weight corresponding to each network edge according to the target behavior data;

[0027] Generate the adaptive spatio-temporal network model according to the network nodes, the network edges, the edge weights, the time attribute, and the space attribute.

[0028] In some embodiments, performing behavioral pattern evolution analysis according to the adaptive spatio-temporal network model to obtain a behavioral pattern evolution analysis result, including:

[0029] Dividing the time series data in the adaptive spatio-temporal network model into time scales to obtain time scale data, where the time scale data includes daily data, weekly data, and monthly data;

[0030] Performing multi-scale analysis on the time scale data using multi-scale time series analysis methods to obtain behavioral patterns at different time scales, where the multi-scale time series analysis methods include wavelet transform or multi-resolution analysis;

[0031] Inputting the behavioral pattern into a graph neural network to perform node representation learning to obtain node vectors;

[0032] Fusing the node vectors with spatio-temporal information using a time convolution or recurrent neural network structure to obtain the time evolution information of the behavioral pattern;

[0033] Performing sequence modeling based on the time evolution information to obtain the dynamic change information of the behavioral pattern;

[0034] Performing relationship reasoning based on the dynamic change information to obtain the change information of the interaction between agents;

[0035] Generating the evolution trend of the behavioral pattern as the behavioral pattern evolution analysis result based on the time evolution information, the dynamic change information, and the change information of the interaction between agents.

[0036] In some embodiments, the method further includes:

[0037] Predicting future behavioral pattern changes using a causal inference model based on the behavioral pattern evolution analysis result;

[0038] Performing risk warning based on the future behavioral pattern changes.

[0039] On the other hand, an embodiment of the present invention provides an adaptive spatio-temporal network behavioral pattern evolution analysis device, including:

[0040] A first module for collecting multi-dimensional behavioral data of agents using machine learning, where the multi-dimensional behavioral data of agents includes financial data, social media data, and news report data;

[0041] A second module for performing data quality monitoring according to the user's analysis target and updating the multi-dimensional behavioral data of agents;

[0042] A third module for performing tensor decomposition on the updated multi-dimensional behavioral data of the agent to obtain target behavioral data;

[0043] A fourth module for constructing an adaptive spatio-temporal network model according to the target behavioral data, where the type of the adaptive spatio-temporal network model includes a directed graph, an undirected graph, or a weighted graph;

[0044] A fifth module for performing behavioral pattern evolution analysis according to the adaptive spatio-temporal network model to obtain a behavioral pattern evolution analysis result.

[0045] On the other hand, an embodiment of the present invention provides a computer device, including:

[0046] At least one processor;

[0047] At least one memory for storing at least one program;

[0048] When the at least one program is executed by the at least one processor, the at least one processor implements the method described above.

[0049] On the other hand, an embodiment of the present invention provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the method described above is implemented.

[0050] The beneficial effects of the present invention are as follows:

[0051] In an embodiment of the present invention, first, machine learning is used to collect multi-dimensional behavioral data of an agent, then data quality monitoring is performed according to the user analysis target, the multi-dimensional behavioral data of the agent is updated, tensor decomposition is performed on the updated multi-dimensional behavioral data of the agent to obtain target behavioral data, then an adaptive spatio-temporal network model is constructed according to the target behavioral data, and finally, behavioral pattern evolution analysis is performed according to the adaptive spatio-temporal network model to obtain a behavioral pattern evolution analysis result, so that behavioral pattern evolution analysis can be realized by using tensor decomposition and an adaptive spatio-temporal network model, thereby improving the analysis accuracy and analysis depth.

[0052] Other features and advantages of the present invention will be described in the following specification, and part of them will be obvious from the specification or understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the specification and the drawings. Description of the Drawings

[0053] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0054] Figure 1 It is a flowchart of a method for analyzing the evolution of an adaptive spatio-temporal network behavior pattern according to an embodiment of the present invention;

[0055] Figure 2 It is a schematic diagram of the overall process of analyzing the evolution of a behavior pattern and early warning according to an embodiment of the present invention;

[0056] Figure 3 It is a schematic structural diagram of an apparatus for analyzing the evolution of an adaptive spatio-temporal network behavior pattern according to an embodiment of the present invention;

[0057] Figure 4 It is a schematic hardware structure diagram of a computer device according to an embodiment of the present invention. Detailed implementation manners

[0058] In order to make the objectives, technical solutions and advantages of the present application clearer, the following further details the present application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. When the following description involves the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application detailed in the appended claims.

[0059] It can be understood that the terms "first", "second", etc. used in the present application can be used in this document to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information can also be called the second information, and similarly, the second information can also be called the first information. Depending on the context, the words "if", "when" as used herein can be interpreted as "when...", "while...", or "in response to determining".

[0060] The terms "at least one", "a plurality", "each", "any one", etc. used in the present application, at least one includes one, two or more than two, a plurality includes two or more than two, each refers to each of the corresponding plurality, and any one refers to any one of the plurality.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terms used herein are for the purpose of describing embodiments of this application only and are not intended to limit this application.

[0062] Before elaborating on the embodiments of this application in detail, some nouns and terms involved in the embodiments of this application are first explained, and the nouns and terms involved in the embodiments of this application are applicable to the following explanations.

[0063] A behavior model is the structured, content-based, and regular series of behaviors of people's daily activities that are motivated, goal-oriented, and characteristic. It is the stereotyped form of behavior content and methods, the "externalization" of life values, and reflects people's action characteristics and behavioral logic. Observed from the perspective of time, a certain behavior model is the program structure of activity time allocation. Observed from the perspective of space, it is the distribution of activity locations and scopes. The specific type to which a person's behavior model belongs is restricted by external environmental conditions, the roles played by the person himself, and life values.

[0064] In related technologies, an agent refers to an entity participating in transactions or interactions in the financial market, including individual investors, institutional investors, algorithmic trading programs, etc. Existing technologies such as time series analysis and clustering analysis have been applied in financial data analysis, but these technologies still fail to fully capture the dynamic evolution and complex network characteristics of agent behavior models. By analyzing historical data to capture agent behavior models, one can only see "what happened in the past", and it is difficult to capture real-time behavior changes, resulting in low dynamicity and low analysis accuracy. At the same time, traditional methods usually only analyze a single agent, but there is a complex relationship network among agents, such as how investors influence each other or the linkage effect between different market segments, resulting in low analysis depth.

[0065] In view of this, the embodiments of the present invention provide an adaptive spatio-temporal network behavior pattern evolution analysis method, device, and storage medium. By performing tensor decomposition on behavior pattern data and using an adaptive spatio-temporal network model for behavior pattern evolution analysis, it provides deeper insights into behavior patterns, like installing a "high-definition dynamic camera" in the financial market, enabling a more in-depth observation and analysis of the dynamic changes and complex relationships of behaviors, thereby improving the analysis accuracy and depth.

[0066] An adaptive spatio-temporal network behavior pattern evolution analysis method provided by an embodiment of the present application relates to the technical field of behavioral data analysis. The adaptive spatio-temporal network behavior pattern evolution analysis method provided by the embodiment of the present application can be applied to a terminal, can also be applied to a server, or can be software running on a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, a vehicle-mounted terminal, etc., but is not limited thereto; the server side can be configured as an independent physical server, can also be configured as a server cluster or a distributed system composed of multiple physical servers, and can also be configured as a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application implementing an adaptive spatio-temporal network behavior pattern evolution analysis method, etc., but is not limited to the above forms.

[0067] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0068] The following specifically explains the embodiments of the present application with reference to the accompanying drawings:

[0069] Figure 1 is an optional flowchart of an adaptive spatio-temporal network behavior pattern evolution analysis method provided by an embodiment of the present application, Figure 1 The method in may include but is not limited to steps S101 to S105.

[0070] Step S101, collecting multi-dimensional behavior data of agents by using machine learning, where the multi-dimensional behavior data of agents includes financial data, social media data, and news report data;

[0071] Step S102: Monitor data quality according to the user's analysis objective, and update the multi-dimensional behavior data of the agent.

[0072] Step S103: Perform tensor decomposition on the updated multi-dimensional behavior data of the agent to obtain the target behavior data.

[0073] Step S104: Construct an adaptive spatio-temporal network model according to the target behavior data. The types of the adaptive spatio-temporal network model include directed graph, undirected graph or weighted graph.

[0074] Step S105: Perform behavior pattern evolution analysis according to the adaptive spatio-temporal network model to obtain the behavior pattern evolution analysis result.

[0075] Steps S101 to S105 shown in the embodiments of the present application achieve behavior pattern evolution analysis, and improve the analysis accuracy and depth.

[0076] In some embodiments, in step S101, using machine learning to collect the multi-dimensional behavior data of the agent may include, but is not limited to, the following steps:

[0077] Use machine learning to identify the importance of the data source.

[0078] Set the data collection frequency and data collection range according to the importance of the data source.

[0079] Collect the multi-dimensional behavior data of the agent according to the data collection frequency and data collection range.

[0080] In some embodiments, the multi-dimensional behavior data of the agent can be collected using machine learning. Among them, the multi-dimensional behavior data of the agent includes financial data, social media data and news report data. First, machine learning can be used to identify the importance of the data source. Exemplarily, machine learning algorithms such as supervised learning or unsupervised learning can be used to identify and predict the importance of the data source and its change trend, so as to automatically identify which data sources and types are most critical for the current market situation. Then, according to the importance of the data source, set the data collection frequency and data collection range, and collect the multi-dimensional behavior data of the agent according to the data collection frequency and data collection range. Exemplarily, the data collection frequency and range can be automatically adjusted according to the output of the machine learning algorithm. For emergencies or abnormal market behaviors, the data collection frequency can be temporarily increased to obtain more detailed data and ensure that key information is not missed. This embodiment maintains a diverse list of data sources, including financial data, social media data, news report data and satellite images, etc.

[0081] In some embodiments, in step S102, according to the user's analysis objective, data quality monitoring is performed to update the multi-dimensional behavior data of the agent, which may include but is not limited to the following steps:

[0082] According to the user's analysis objective, perform data quality analysis on the multi-dimensional behavior data of the agent to obtain a data quality analysis result. Data quality analysis includes integrity analysis, accuracy analysis, consistency analysis, and timeliness analysis;

[0083] If the data quality analysis result is that the quality does not meet the standard, re-collect data from the data backup source and update the multi-dimensional behavior data of the agent.

[0084] In some embodiments, first, according to the user's analysis objective, perform data quality analysis on the multi-dimensional behavior data of the agent to obtain a data quality analysis result. Among them, data quality analysis includes integrity analysis, accuracy analysis, consistency analysis, and timeliness analysis. Exemplarily, according to the user's analysis objective, adaptively select and prioritize data sources, and monitor data quality in real time, such as integrity, accuracy, consistency, and timeliness. If the data quality analysis result is that the quality does not meet the standard, or it is detected that the data quality has decreased, then automatically take measures to re-collect data from the data backup source and update the multi-dimensional behavior data of the agent. It is also possible to re-collect data from the original data source, and it is also possible to verify and correct the data. It can be understood that the user's analysis objective refers to the specific information or result that the user wants to obtain through system analysis. The user's analysis objective determines which data should be collected and how to process these data. For example, if the user wants to analyze the investment trends in a certain stock market, then the analysis objective may be: identify the main investor groups in the market, analyze the trading behavior patterns of different investor groups, and predict future market trends. Data such as stock trading data, investor information, and market news can be collected, and then a spatio-temporal network model can be constructed based on these data for multi-scale evolution analysis, and finally the analysis result can be obtained. More importantly, a feedback loop can be established to feedback the results of subsequent data analysis to the data collection process to optimize future data collection strategies, so as to automatically adjust the data collection process, ensure that high-quality and highly relevant data can be continuously provided, and thus support the effective conduct of behavior pattern evolution analysis.

[0085] In some embodiments, in step S103, perform tensor decomposition on the updated multi-dimensional behavior data of the agent to obtain target behavior data, which may include but is not limited to the following steps:

[0086] Extract features from the updated multi-dimensional behavior data of the agent to obtain a core tensor;

[0087] Determine the data dimension according to the updated multi-dimensional behavior data of the agent;

[0088] Set the factor matrix corresponding to each dimension according to the data dimension;

[0089] Generate a reconstructed tensor as the target behavior data based on the core tensor and multiple factor matrices.

[0090] In some embodiments, in financial data analysis, data is usually stored in tabular form, and these tables can be regarded as two-dimensional matrices. However, when the data has multiple dimensions (such as time, product, region, user, etc.), traditional tables are difficult to process. The complex data can be split into simple modules through tensor decomposition and feature extraction to reduce the data dimension and extract key features. First, feature extraction can be performed on the updated multi-dimensional behavior data of the agent to obtain the core tensor, and based on the updated multi-dimensional behavior data of the agent, the data dimension can be determined. Exemplarily, when there is a very small financial data set, it is a three-dimensional tensor representing the investment amounts of two investors (Investor 1 and Investor 2) in three different financial products (Stock A, Stock B, Bond C) in three different months (January, February, March). The dimension of this three-dimensional tensor is 3x2x3 (month x investor x product). Feature extraction is performed on the updated multi-dimensional behavior data of the agent to obtain a smaller tensor as the core tensor, which will capture the main features of the data. In this embodiment, it can be assumed that the core tensor is a 1x1x1 tensor, and its value is the average of all investment amounts. If the average of all investment amounts is 2500, then the core tensor G is: G = 2500.

[0091] Then, according to the data dimension, set the factor matrix corresponding to each dimension, and generate a reconstructed tensor as the target behavior data based on the core tensor and multiple factor matrices. Exemplarily, a factor matrix can be found for each dimension, and these matrices will capture the features on each dimension. The factor matrix A corresponding to the month dimension can be set, and a 3x1 matrix can be selected to represent the features of the month dimension, which represents the weights of January, February, and March. The factor matrix B corresponding to the investor dimension can be set, and a 2x1 matrix can be selected to represent the features of the investor dimension, which represents the investment amount ratios of Investor 1 and Investor 2. The factor matrix C corresponding to the product dimension can be set, and a 3x1 matrix can be selected to represent the features of the product dimension, which represents the investment amount ratios of Stock A, Stock B, and Bond C. Then use these factor matrices and the core tensor to reconstruct the original tensor. The reconstruction process is completed by calculating the outer product of the factor matrices and the product of the core tensor. Calculate the outer product and the final product to obtain a reconstructed tensor T' close to the original tensor. The calculation formula for the reconstructed tensor is: T' = G x1 A x2 B x3 C, where x1, x2, x3 all represent the operators of the tensor outer product, A, B, C all represent the factor matrices, G represents the core tensor, and T' represents the reconstructed tensor.

[0092] It is understandable that the outer product of tensors is an operation that combines two tensors into a higher-dimensional tensor. Three two-dimensional factor matrices (A, B, C) and a one-dimensional core tensor G are combined into a three-dimensional tensor T. The combined structure is simpler, and it is no longer necessary to process all the data in the original three-dimensional tensor T. Instead, only a few tensor combination data need to be processed, reducing the dimension. At the same time, the main features of the data are extracted and concentrated in the core tensor G. And redundant information is removed, only the most important information is retained, making the information concentrated. In this embodiment, a three-dimensional financial data tensor is decomposed into the product of a core tensor and three factor matrices, thereby reducing the dimension of the data and extracting key features.

[0093] Furthermore, feature extraction refers to the process of identifying and selecting the most useful information (features) from the original data. These features are crucial for subsequent data analysis and modeling. Methods of feature extraction include principal component analysis (PCA) or factor analysis, etc., which are used to identify the main variation directions in the data. Information concentration refers to extracting a few features that can best represent the essence of the data from a large amount of original data. Model performance enhancement refers to improving the accuracy and generalization ability of the model by removing noise and irrelevant information. Tensor decomposition is a method used to process high-dimensional data. It reduces the dimension of the data through decomposition techniques while retaining the main structure of the data. By decomposing a tensor into the product of a core tensor and multiple factor matrices, these factor matrices correspond to different dimensions of the tensor respectively.

[0094] In some embodiments, in step S104, according to the target behavior data, constructing an adaptive spatio-temporal network model may include, but is not limited to, the following steps:

[0095] According to the target behavior data, set network nodes, and the network nodes include agents or agent behaviors;

[0096] According to the target behavior data and network nodes, set network edges, and the network edges are used to represent the associations between network nodes;

[0097] According to the target behavior data, set the edge weight corresponding to each network edge;

[0098] According to the network nodes, network edges, edge weights, time attributes, and space attributes, generate an adaptive spatio-temporal network model.

[0099] In some embodiments, an adaptive spatio-temporal network model can be constructed based on target behavior data, and the network structure can be dynamically adjusted. Among them, the types of adaptive spatio-temporal network models include directed graphs, undirected graphs, or weighted graphs. First, network nodes can be set according to the target behavior data. Among them, network nodes include agents or agent behaviors. Exemplarily, an agent (such as an investor, a trader) or an agent behavior (such as a trading behavior) can be defined as a network node. Then, according to the target behavior data and the network nodes, network edges are set, where the network edges are used to represent the associations between network nodes. Exemplarily, edges between nodes, that is, network edges, can be established according to the interactions between agents (such as trading relationships, information dissemination) or the associations between behaviors (such as similar investment strategies). Then, according to the target behavior data, the edge weight corresponding to each network edge is set. The edge weight can represent the intensity or frequency of the interaction. For example, the size of the trading amount or the frequency of transactions. Finally, an adaptive spatio-temporal network model is generated according to the network nodes, network edges, edge weights, time attributes, and spatial attributes. Exemplarily, the time attribute can include a trading time series, the spatial attribute can include a market sector or a geographical location, and the adaptive spatio-temporal network model can be dynamically updated as new data arrives. Further, the adaptive spatio-temporal network model is a complex network model that combines time and space dimensions to represent and analyze the behaviors of agents. The types of the model can include directed graphs, undirected graphs, or weighted graphs. A directed graph indicates that the interaction between agents is directional, such as information transmission. An undirected graph indicates that the interaction is two-way, such as co-investment. A weighted graph indicates that the intensity or importance of the interaction needs to be represented. The adaptive spatio-temporal network model contains time series data of agent behaviors and their spatial relationships. The network state at each time point is regarded as a snapshot, and these snapshots are arranged in chronological order to form a dynamic network evolution process.

[0100] In some embodiments, in step S105, according to the adaptive spatio-temporal network model, behavioral pattern evolution analysis is performed to obtain a behavioral pattern evolution analysis result, which may include but is not limited to the following steps:

[0101] Perform time scale division on the time series data in the adaptive spatio-temporal network model to obtain time scale data, where the time scale data includes daily data, weekly data, and monthly data;

[0102] Use multi-scale time series analysis methods to perform multi-scale analysis on the time scale data to obtain behavioral patterns at different time scales. The multi-scale time series analysis methods include wavelet transform or multi-resolution analysis;

[0103] Input the behavioral pattern into a graph neural network to perform node representation learning to obtain node vectors;

[0104] Fusing the node vectors with spatio-temporal information using a temporal convolutional or recurrent neural network structure to obtain the temporal evolution information of the behavior pattern;

[0105] Performing sequence modeling based on the temporal evolution information to obtain the dynamic change information of the behavior pattern;

[0106] Performing relationship reasoning based on the dynamic change information to obtain the change information of the interaction between agents;

[0107] Generating the evolution trend of the behavior pattern as the result of the behavior pattern evolution analysis based on the temporal evolution information, the dynamic change information, and the change information of the interaction between agents.

[0108] In some embodiments, an adaptive spatio-temporal network model can be used, combined with multi-scale time series analysis method and graph neural network technology, to analyze the evolution trend of the behavior pattern. The time series data in the adaptive spatio-temporal network model can be first divided into time scales to obtain time scale data, which can capture the behavior patterns at different time scales. Among them, the time scale data includes daily data, weekly data, and monthly data. Then, the multi-scale time series analysis method is used to perform multi-scale analysis on the time scale data to obtain the behavior patterns at different time scales, so as to identify the trends and periodicities of the behavior patterns at different time scales, and compare the behavior patterns at different time scales to identify the persistence and change points of the patterns, enabling the observation of short-term (such as daily) and long-term (such as monthly) behavior changes at the same time. Among them, the multi-scale time series analysis method includes wavelet transform or multi-resolution analysis. It can be understood that the multi-scale time series analysis method can analyze the behavior patterns at different time scales, such as daily, weekly, monthly, etc. Each analysis at a time scale will obtain a result, so multiple results will be obtained finally, and these results are used as the input data of the graph neural network technology. Through multi-scale time series analysis, the change trends of the behavior pattern, such as growth, decrease, periodic fluctuations, etc., can be detected.

[0109] Exemplarily, assume that it is necessary to analyze the trading behavior of investors in the stock market, and the multi-scale time series analysis method is used to observe the trading patterns at different time scales. In the daily analysis, it is observed that in the past few days, the trading frequency of investor A has gradually increased, which may indicate that investor A has specific short-term expectations for the market. In the weekly analysis, on the weekly time scale, it is observed that investor A is more active in trading on specific days of each week (such as Tuesday and Thursday). In the monthly analysis, from the monthly data, it is observed that the trading volume of investor A significantly increases in specific months (such as the end of the year), which may be related to his personal financial planning.

[0110] Then, input the behavior patterns into a graph neural network for node representation learning to obtain node vectors. Exemplarily, a graph neural network (GNN) can be used to learn the representations of each node in the network, and these representations can capture the neighborhood information of the nodes, i.e., the interactions between agents. Use a temporal convolutional or recurrent neural network structure to fuse the node vectors with spatio-temporal information to obtain the temporal evolution information of the behavior patterns. Exemplarily, a GNN can process dynamic networks and fuse spatio-temporal information through a temporal convolutional or recurrent neural network structure to learn the evolution of behavior patterns over time. The spatio-temporal information can include time information and space information. According to the temporal evolution information, perform sequence modeling to obtain the dynamic change information of the behavior patterns. Exemplarily, the sequence modeling ability of a GNN can be used to analyze the evolution of node states over time to capture the dynamic changes of behavior patterns. According to the dynamic change information, perform relational reasoning to obtain the change information of the interactions between agents. Exemplarily, through the relational reasoning ability of a GNN, analyze the changes in the interactions between agents and how these changes affect the behavior patterns of the entire network. Finally, based on the temporal evolution information, the dynamic change information, and the change information of the interactions between agents, generate the evolution trend of the behavior patterns as the result of the behavior pattern evolution analysis. It can be understood that using the classification ability of a GNN, the behavior patterns of agents can be classified into different categories, such as stable, fluctuating, abnormal, etc. Node representation learning is to create a node vector for each node. Spatio-temporal information fusion is to combine these node vectors with time and space information. Sequence modeling is to analyze the changes of these information over time, and relational reasoning is to use these information to understand the interactions between nodes. The relationship between node representation learning, spatio-temporal information fusion, sequence modeling, and relational reasoning is progressive, and each step provides the necessary information for the next step. Further, by analyzing the network structure, the key behavior patterns of agents can be identified, such as the emergence of leaders, the formation of groups, or the detection of abnormal behaviors. It can also help predict market trends by analyzing the evolution of behavior patterns in the network.

[0111] Exemplarily, graph neural network technology can be used to analyze the trading network among investors. In the network structure, a graph can be constructed, where the network nodes represent investors, the network edges represent the trading relationships among investors, and the edge weights represent the trading amounts. In node representation learning, the graph neural network captures the interaction patterns between investor A and his trading partners by learning the representations of each network node. In spatio-temporal information fusion, the graph neural network considers time series data, enabling it to learn how investor A's trading behavior evolves over time and how this evolution affects the entire network. The evolution trend is the comprehensive result of the analysis of the evolution of behavior patterns, revealing the changing patterns of investors' behaviors. In the results of the analysis of pattern evolution, in the short-term trend analysis results, the increase in investor A's daily trading frequency may indicate that he is sensitive to short-term market fluctuations. In the medium-term trend analysis results, the trading patterns shown in the weekly analysis may reflect investor A's work schedule or market participation habits. In the long-term trend analysis results, the annual trading patterns revealed by the monthly analysis may be related to investor A's long-term investment strategy.

[0112] In some embodiments, the method further includes:

[0113] Predicting future changes in behavior patterns using a causal inference model based on the results of the analysis of the evolution of behavior patterns;

[0114] Performing risk warnings based on the future changes in behavior patterns.

[0115] In some embodiments, first, based on the results of the analysis of the evolution of behavior patterns, a causal inference model can be used to predict future changes in behavior patterns. Exemplarily, through the causal inference model, market trends can be predicted, such as analyzing how policy changes affect investors' behaviors. Then, risk warnings are performed based on the future changes in behavior patterns, such as detecting abnormal selling by multiple investors and issuing risk warnings in advance. More specifically, risk monitoring can be carried out. The adaptive spatio-temporal network model can be used to monitor the spread of risks in the market, such as the spread of financial crises. It can also provide decision support for investors by identifying key nodes and potential investment opportunities in the network.

[0116] In some embodiments, the overall process of performing the analysis of the evolution of behavior patterns and warnings is as Figure 2 shown. First, adaptive data collection can be performed, including real-time data collection, dynamically adjusting data sources, monitoring data quality, and feedback for optimizing the collection strategy. Then, high-dimensional data processing is carried out, including tensor decomposition, feature extraction, and reducing the data dimension. Next, an adaptive spatio-temporal network model is constructed, including defining nodes and edges, assigning weights, and dynamically updating the network structure. Then, the analysis of the evolution of behavior patterns is performed, including multi-scale time series analysis, graph neural network analysis, and detecting the evolution of behavior patterns. Finally, intelligent warnings and decision support are provided, including predicting future behavior patterns, risk warnings, and providing decision-making suggestions.

[0117] In some embodiments, this embodiment realizes the in-depth tracking and analysis of the evolution of the agent's behavior pattern through an innovative adaptive spatio-temporal network model construction, multi-scale time series analysis method, and causal inference model. The analysis results of the behavior pattern evolution can not only provide more accurate and forward-looking analysis for the financial market, but also effectively predict and prevent market risks, having important application value and market potential.

[0118] The beneficial effects of implementing the embodiments of the present invention include: The embodiments of the present invention first use machine learning to collect multi-dimensional behavior data of the agent, then perform data quality monitoring according to the user's analysis target, update the multi-dimensional behavior data of the agent, perform tensor decomposition on the updated multi-dimensional behavior data of the agent to obtain target behavior data, then construct an adaptive spatio-temporal network model according to the target behavior data, and finally perform behavior pattern evolution analysis according to the adaptive spatio-temporal network model to obtain the analysis results of the behavior pattern evolution, so as to be able to use tensor decomposition and the adaptive spatio-temporal network model to realize behavior pattern evolution analysis, thereby improving the analysis accuracy and depth.

[0119] As Figure 3 shown, the embodiments of the present invention also provide an adaptive spatio-temporal network behavior pattern evolution analysis device, including:

[0120] The first module 801 is used to collect multi-dimensional behavior data of the agent by using machine learning, and the multi-dimensional behavior data of the agent includes financial data, social media data, and news report data;

[0121] The second module 802 is used to perform data quality monitoring according to the user's analysis target and update the multi-dimensional behavior data of the agent;

[0122] The third module 803 is used to perform tensor decomposition on the updated multi-dimensional behavior data of the agent to obtain target behavior data;

[0123] The fourth module 804 is used to construct an adaptive spatio-temporal network model according to the target behavior data, and the types of the adaptive spatio-temporal network model include directed graphs, undirected graphs, or weighted graphs;

[0124] The fifth module 805 is used to perform behavior pattern evolution analysis according to the adaptive spatio-temporal network model to obtain the analysis results of the behavior pattern evolution.

[0125] The content in the above method embodiments is applicable to the device embodiments of the present invention. The functions specifically implemented by the device embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0126] As Figure 4As shown, an embodiment of the present invention further provides a computer device, including:

[0127] At least one processor 901;

[0128] At least one memory 902 for storing at least one program;

[0129] When the at least one program is executed by the at least one processor, the at least one processor implements Figure 1 The method shown.

[0130] The content in the above method embodiments is applicable to the device embodiments of the present invention. The functions specifically implemented by the device embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0131] An embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements Figure 1 The method shown.

[0132] The content in the above method embodiments is applicable to the storage medium embodiments of the present invention. The functions specifically implemented by the storage medium embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0133] The preferred embodiments of the embodiments of the present application have been described above with reference to the accompanying drawings. However, this does not limit the scope of the rights of the embodiments of the present application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall fall within the scope of the rights of the embodiments of the present application.

Claims

1. An analysis method for the evolution of adaptive spatio-temporal network behavior patterns, characterized in that, Including the following steps: Collecting multi-dimensional behavior data of agents using machine learning, where the multi-dimensional behavior data of agents includes financial data, social media data, and news report data; According to the user analysis objective, conducting data quality monitoring and updating the multi-dimensional behavior data of agents; Performing tensor decomposition on the updated multi-dimensional behavior data of agents to obtain target behavior data; Constructing an adaptive spatio-temporal network model based on the target behavior data, and the types of the adaptive spatio-temporal network model include directed graph, undirected graph, or weighted graph; Conducting behavior pattern evolution analysis based on the adaptive spatio-temporal network model to obtain the behavior pattern evolution analysis result.

2. The method according to claim 1, wherein The collecting multi-dimensional behavior data of agents using machine learning includes: Identifying the importance degree of data sources using machine learning; Setting the data collection frequency and data collection range according to the importance degree of the data sources; Collecting the multi-dimensional behavior data of agents according to the data collection frequency and the data collection range.

3. The method according to claim 1, wherein The conducting data quality monitoring and updating the multi-dimensional behavior data of agents according to the user analysis objective includes: Conducting data quality analysis on the multi-dimensional behavior data of agents according to the user analysis objective to obtain the data quality analysis result, and the data quality analysis includes integrity analysis, accuracy analysis, consistency analysis, and timeliness analysis; If the data quality analysis result is that the quality does not meet the standard, re-collecting data from the data backup source and updating the multi-dimensional behavior data of agents.

4. The method according to claim 1, wherein The performing tensor decomposition on the updated multi-dimensional behavior data of agents to obtain target behavior data includes: Performing feature extraction on the updated multi-dimensional behavior data of agents to obtain the core tensor; Determining the data dimension according to the updated multi-dimensional behavior data of agents; Setting the factor matrix corresponding to each dimension according to the data dimension; Generating a reconstructed tensor as the target behavior data according to the core tensor and multiple factor matrices.

5. The method according to claim 1, characterized in that The constructing an adaptive spatio-temporal network model based on the target behavior data includes: Setting network nodes according to the target behavior data, where the network nodes include agents or agent behaviors; Setting network edges according to the target behavior data and the network nodes, and the network edges are used to represent the associations between the network nodes; Setting the edge weight corresponding to each network edge according to the target behavior data; Generating the adaptive spatio-temporal network model according to the network nodes, the network edges, the edge weights, time attributes, and space attributes.

6. The method according to claim 1, wherein The conducting behavior pattern evolution analysis based on the adaptive spatio-temporal network model to obtain the behavior pattern evolution analysis result includes: Conducting time scale division on the time series data in the adaptive spatio-temporal network model to obtain time scale data, where the time scale data includes daily data, weekly data, and monthly data; Performing multi-scale analysis on the time scale data using multi-scale time series analysis methods to obtain behavior patterns at different time scales, and the multi-scale time series analysis methods include wavelet transform or multi-resolution analysis; Input the behavioral pattern into a graph neural network for node representation learning to obtain node vectors; Use a temporal convolutional or recurrent neural network structure to fuse the node vectors with spatio-temporal information to obtain the temporal evolution information of the behavioral pattern; Perform sequence modeling based on the temporal evolution information to obtain the dynamic change information of the behavioral pattern; Perform relational reasoning based on the dynamic change information to obtain the change information of the interaction between agents; Generate the evolution trend of the behavioral pattern as the analysis result of the behavioral pattern evolution based on the temporal evolution information, the dynamic change information, and the change information of the interaction between agents; 7. The method according to claim 1, wherein The method further includes: Predict the future change of the behavioral pattern using a causal inference model based on the analysis result of the behavioral pattern evolution; Perform risk warning based on the future change of the behavioral pattern; 8. An adaptive spatio-temporal network behavior pattern evolution analysis device, characterized in that It includes: A first module for collecting multi-dimensional behavioral data of agents using machine learning, where the multi-dimensional behavioral data of agents includes financial data, social media data, and news report data; A second module for monitoring data quality according to the user's analysis target and updating the multi-dimensional behavioral data of agents; A third module for performing tensor decomposition on the updated multi-dimensional behavioral data of agents to obtain target behavioral data; A fourth module for constructing an adaptive spatio-temporal network model based on the target behavioral data, where the type of the adaptive spatio-temporal network model includes a directed graph, an undirected graph, or a weighted graph; A fifth module for performing behavioral pattern evolution analysis based on the adaptive spatio-temporal network model to obtain the analysis result of the behavioral pattern evolution; 9. A computer device, characterized in that, It includes: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1-7; 10. A computer-readable storage medium storing a computer program, characterized in that, The computer program, when executed by a processor, implements the method according to any one of claims 1-7.

Images