A digital asset user analysis encryption method, system and readable storage medium

Abstract

The application discloses a kind of digital asset user analysis encryption method, system and readable storage medium.By using the multi-source heterogeneous data analysis for behavior data and scene data, important support is provided for the identification and selection of user information, accurate portrait, which can improve customer satisfaction.Through the analysis of the emotional characteristics of network individuals and groups, and the extraction of behavior and scene information characteristics such as text feedback, the application can locate core value customers.With customer behavior data, feature data, social interaction data, etc.as input dimensions, while analyzing the customer preference, the user's social influence and social network can also be analyzed, and then the user's social status, economic conditions, education level, etc.are studied.

Description

Technical Field

[0001] This invention relates to the field of data analysis and processing, and more specifically, to a digital asset user analysis encryption method, system, and readable storage medium. Background Technology

[0002] Several works have already analyzed digital assets from the perspectives of security, privacy, applications, usability, regulations, and technology. While there are text mining analyses specifically for profiling digital asset activity, no work has yet been proposed that focuses on text mining for price prediction and on analyzing social network data to better understand the underlying factors influencing prices and other trends in emerging digital asset markets. Furthermore, customer profiling could be conducted based on research into social network data to understand the relationship between digital asset prices and social activity. Keywords could be extracted from user comments related to digital assets posted on online forums to analyze and predict transaction prices and ranges, as well as digital asset price volatility. However, previous work has primarily relied on text analysis to predict digital assets.

[0003] Therefore, it is crucial to analyze and identify digital asset investors and those interested in them to protect new investors from financial losses and to provide information to new investors and interested parties. Existing technologies do not consider analyzing users' social networks and lack detailed analysis of the richness of attributes within the network, nor do they consider analysis of active user characteristics based on balance thresholds. Summary of the Invention

[0004] In view of the above problems, the purpose of this invention is to provide a digital asset user analysis encryption method, system and readable storage medium. By using multi-source heterogeneous data analysis of behavioral data and scenario data, it provides important support for user information identification and selection and accurate profiling, which can improve customer satisfaction.

[0005] The first aspect of this invention provides a digital asset user analysis encryption method, comprising the following steps:

[0006] Extract structured information from text documents containing behavioral and contextual data from social media and online platforms into comma-separated value (CSV) files;

[0007] The structured results processed by the text analysis platform are combined with other structured data sources, and the dimensional data is transformed and loaded into the underlying data warehouse using SNM's data integration tool (Zanho).

[0008] Integrated preprocessing of multi-source heterogeneous data in a data warehouse;

[0009] This system utilizes an artificial intelligence engine based on natural language processing / language engineering to analyze and predict pre-processed multi-source heterogeneous data, and outputs the analysis and prediction results.

[0010] In this solution, the integrated preprocessing of multi-source heterogeneous data in the data warehouse includes:

[0011] Construct heterogeneous graph convolutional networks for multi-source heterogeneous data in data warehouses;

[0012] A novel two-layer attention mechanism is used to preprocess behavioral and scene data.

[0013] In this scheme, the construction of the heterogeneous graph convolutional network specifically involves:

[0014] Define an undirected graph G=(V,E;O,S) that uses entity type mapping:φ:V→O and link type mapping:∅:E→S; vertex υ∈V is an entity, and edge e= = V,u>∈e represents the relationship between V and u, where υ,u∈V;

[0015] ti∈O={o1,o2,…,on} is an entity type, φ(υ)∈T;

[0016] sj∈S={s1,s2,…,sm} is a relation, ∅(e)∈R; the number of entity types |M|>1 and the number of relation types |R|>1; all vertices V={V1∪V2,…∪Vn} can be partitioned into n mutually exclusive subsets; all edges E={E1∪E2,…∪Em} can be partitioned into m mutually exclusive subsets;

[0017] In a weighted network, w(e) represents the weight of e = ⟨v, u>∈e.

[0018] In this scheme, the novel two-layer attention mechanism includes a node-level attention machine and a type-level attention.

[0019] In this scheme, the novel dual-layer attention mechanism is used to preprocess behavioral and scene data. Specifically, HIN (Hetegeneous Information Network) is embedded into HGCN (Heterogeneous Graph Convolutional Network Deep Learning Model for Collective Classification), and then a dual-layer attention mechanism is introduced to calculate the attention weights of different neighboring nodes and nodes of different types.

[0020] In this solution, the specific steps of using an artificial intelligence engine based on natural language processing / language engineering to analyze and predict preprocessed multi-source heterogeneous data are as follows:

[0021] First, collect several days' worth of data and store it in an Apache Spark RDD, then perform computations to build features from the raw data to guide the AI ​​engine;

[0022] Next, a data stream is initiated to obtain behavioral and contextual data and related real-time data from social media; features are calculated on this data to obtain new data points for future predictions from the artificial intelligence engine.

[0023] A second aspect of the present invention provides a digital asset user analysis encryption system, comprising a memory and a processor. The memory includes a digital asset user analysis encryption method program, which, when executed by the processor, performs the following steps:

[0024] Extract structured information from text documents containing behavioral and contextual data from social media and online platforms into comma-separated value (CSV) files;

[0025] The structured results processed by the text analysis platform are combined with other structured data sources, and the dimensional data is transformed and loaded into the underlying data warehouse using SNM's data integration tool (Zanho).

[0026] Integrated preprocessing of multi-source heterogeneous data in a data warehouse;

[0027] This system utilizes an artificial intelligence engine based on natural language processing / language engineering to analyze and predict pre-processed multi-source heterogeneous data, and outputs the analysis and prediction results.

[0028] In this solution, the integrated preprocessing of multi-source heterogeneous data in the data warehouse includes:

[0029] Construct heterogeneous graph convolutional networks for multi-source heterogeneous data in data warehouses;

[0030] A novel two-layer attention mechanism is used to preprocess behavioral and scene data.

[0031] In this solution, the specific steps of using an artificial intelligence engine based on natural language processing / language engineering to analyze and predict preprocessed multi-source heterogeneous data are as follows:

[0032] First, collect several days' worth of data and store it in an Apache Spark RDD, then perform computations to build features from the raw data to guide the AI ​​engine;

[0033] Next, a data stream is initiated to obtain behavioral and contextual data and related real-time data from social media; features are calculated on this data to obtain new data points for future predictions from the artificial intelligence engine.

[0034] A third aspect of the present invention provides a computer-readable storage medium comprising a digital asset user analysis encryption method program, wherein when executed by a processor, the digital asset user analysis encryption method program implements the steps of the digital asset user analysis encryption method as described in any of the preceding claims.

[0035] This invention provides crucial support for user information identification and selection, and precise profiling through the analysis of multi-source heterogeneous data targeting behavioral and contextual data, thereby improving customer satisfaction. By extracting behavioral and contextual information features such as emotional characteristics and textual feedback from individuals and groups online, this invention can identify core value customers. Using customer behavior data, feature data, and social interaction data as input dimensions, it can analyze customer preferences, as well as users' social influence and social networks, and further conduct research based on factors such as users' social status, economic conditions, and education level.

[0036] This invention presents an online multi-source heterogeneous data extraction and unstructured data warehouse design model. It helps identify and extract text documents containing behavioral and contextual data from multiple social media and online platforms, and further extracts relevant structured items. This improves recognition efficiency while reducing information loss.

[0037] This invention proposes a novel two-layer attention mechanism based on heterogeneous graph convolutional networks for data analysis of behavioral and scene data. This mechanism can solve the problems of semantic sparsity and hidden network structure information, and can integrate additional information, thereby greatly enriching semantic information.

[0038] The engine described herein, based on natural language processing / language engineering, can scale and manage large amounts of heterogeneous data and visualize the trends of the results. Attached Figure Description

[0039] Figure 1 A flowchart of a digital asset user analysis encryption method according to this application is shown;

[0040] Figure 2 A block diagram of a digital asset user analysis encryption system according to the present invention is shown. Detailed Implementation

[0041] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0042] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0043] Figure 1 A flowchart of a digital asset user analysis encryption method according to this application is shown.

[0044] like Figure 1 As shown, this application discloses a digital asset user analysis encryption method, including the following steps:

[0045] S102: Extract structured information from text documents containing behavioral and contextual data from social media and online platforms into comma-separated value (CSV) files;

[0046] S104: Combine the structured results processed by the text analysis platform with other structured data sources, and use SNM's data integration tool (Zanho) to transform and load the dimensional data into the underlying data warehouse;

[0047] S106: Integrated preprocessing of multi-source heterogeneous data in a data warehouse;

[0048] S108: Utilizes an artificial intelligence engine based on natural language processing / language engineering to analyze and predict pre-processed multi-source heterogeneous data, and outputs the analysis and prediction results.

[0049] It should be noted that in S102, a text analysis platform is used to extract structured information from text documents into comma-separated value (CSV) files. This structured result is then combined with other structured data sources. During the ETL process (transforming, cleaning, and loading the extracted text as a CSV file into the data warehouse), SNM's data integration tool (Zanho) loads it into the underlying data warehouse server. Fact data is transformed, mapped to dimensional data, and loaded into the data warehouse. The import of new data creates new tables for dimensions and facts. The import of incremental data updates or refreshes existing dimensional and fact tables.

[0050] It is worth mentioning the online multi-source heterogeneous data extraction and unstructured data warehouse design model for Social Networks Mining (SNM). This invention combines a text processing engine and a predefined extractor library to help identify and extract text documents containing behavioral and contextual data from multiple social media and online platforms, and further extract relevant structured items from them.

[0051] According to embodiments of the present invention, integrated preprocessing of multi-source heterogeneous data in a data warehouse includes:

[0052] Construct heterogeneous graph convolutional networks for multi-source heterogeneous data in data warehouses;

[0053] A novel two-layer attention mechanism is used to preprocess behavioral and scene data.

[0054] According to an embodiment of the present invention, the construction of the heterogeneous graph convolutional network specifically involves:

[0055] Define an undirected graph G=(V,E;O,S) that uses entity type mapping:φ:V→O and link type mapping:∅:E→S; vertex υ∈V is an entity, and edge e= = V,u>∈e represents the relationship between V and u, where υ,u∈V;

[0056] ti∈O={o1,o2,…,on} is an entity type, φ(υ)∈T;

[0057] sj∈S={s1,s2,…,sm} is a relation, ∅(e)∈R; the number of entity types |M|>1 and the number of relation types |R|>1; all vertices V={V1∪V2,…∪Vn} can be partitioned into n mutually exclusive subsets; all edges E={E1∪E2,…∪Em} can be partitioned into m mutually exclusive subsets;

[0058] In a weighted network, w(e) represents the weight of e = ⟨v, u>∈e.

[0059] According to an embodiment of the present invention, the novel two-layer attention mechanism includes a node-level attention machine and a type-level attention.

[0060] According to an embodiment of the present invention, the preprocessing of behavioral data and scene data using a novel dual-layer attention mechanism specifically involves: embedding HIN (Hetegeneous Information Network) into HGCN (Heterogeneous Graph Convolutional Network Deep Learning Model for Collective Classification), and then introducing a dual-layer attention mechanism to calculate the attention weights of different neighboring nodes and nodes of different types.

[0061] It should be noted that the novel two-layer attention mechanism based on heterogeneous graph convolutional networks is used for data analysis based on behavioral and scene data. It includes node-level attention and type-level attention. We embed HIN (Hetegeneous Information Network) into HGCN (Hypergeneous Graph Convolutional Network Deep Learning Model for Collective Classification), and then introduce the aforementioned two-layer attention mechanism to calculate the attention weights of different neighboring nodes and nodes of different types.

[0062]

[0063] This algorithm can solve the problems of semantic sparsity and hidden network structure information. It can also integrate several additional pieces of information, which can greatly enrich the semantic information.

[0064] According to an embodiment of the present invention, the analysis and prediction of preprocessed multi-source heterogeneous data using an artificial intelligence engine based on natural language processing / language engineering specifically involves:

[0065] First, collect several days' worth of data and store it in an Apache Spark RDD, then perform computations to build features from the raw data to guide the AI ​​engine;

[0066] Next, a data stream is initiated to obtain behavioral and contextual data and related real-time data from social media; features are calculated on this data to obtain new data points for future predictions from the artificial intelligence engine.

[0067] It should be noted that the natural language processing / language engineering-based engine described in this invention aims to analyze and predict using previously collected and preprocessed behavioral and contextual data. This process is achieved by learning the correlation between relevant behavioral and contextual data sentiment and the relevant digital asset users.

[0068] The engine first learns from a given history and then begins making predictions based on previous correlations. It is also able to improve itself through its mistakes and attempts to refine its predictions. Furthermore, the engine supports trend visualization of records over time based on incoming data and intermediate results. Built with the ever-increasing volume, speed, and diversity of available data in mind, the engine is capable of scaling and managing large amounts of heterogeneous data.

[0069] This engine is built within the Apache ecosystem and uses Apache Spark. Spark's data structures are based on Resilient Distributed Datasets (RDDs), which are read-only, multi-set datasets that can be distributed across a group of machines and offer some fault tolerance. Spark applications run as separate processes on different clusters and are coordinated by Spark objects (also known as SparkContexts). This element is the main driver of the program; it connects to the cluster manager and helps acquire executors on different nodes to allocate resources across the application. Spark is highly scalable, up to 100 times faster than Hadoop on large datasets, and provides out-of-the-box libraries for streaming and machine learning. At the heart of the architecture is Apache Spark, which acts as an in-memory data store and allows us to perform computations in a scalable manner. This data is the input to our machine learning model for making predictions. To bootstrap our model, we first collect several days' worth of data and store it in an Apache Spark RDD. Next, we perform computations to build features from the raw data. All of these computations are performed on data distributed across multiple Spark clusters and therefore scale as the data grows.

[0070] Once the AI ​​engine is bootstrapping, we begin a data stream to acquire behavioral and contextual data from social media and related real-time data. Similar computations are performed on this data to compute features, and this new data point is then used to derive future predictions from the model. This computed data point is then appended to the existing data in the Spark RDD, derived from the bootstrap data. Therefore, in addition to the ability to create user profiles, we are continuously expanding our data storage, enabling us to extract holistic visualizations from data on digital asset market trends.

[0071] It is worth mentioning that the core idea of ​​the method described in this invention can be summarized as a novel hardware algorithm for homomorphic encryption operations, which overcomes the need for dedicated memory to store the similarity matrix, allows more processing elements to be placed on the FPGA, achieves high speed and processing complexity, and provides hardware and software benefits, namely flexibility, improved and enhanced speed.

[0072] The complete implementation of this invention relies on a hardware / software co-design method. By sending the subpolynomials to the FPGA, hardware computation completely hides the transmission latency between software and hardware. The accelerator can operate in two modes: a standard mode for polynomial multiplication and a relinearization / key switching mode.

[0073] This invention proposes three hardware implementations of Karatsuba: a region-efficient design with three recursions in the hardware, called Karatsuba-3; a balanced implementation with four recursions in the hardware, called Karatsuba-4; and a performance-oriented design with five recursions in the hardware, called Karatsuba-5.

[0074] For the first design with three recursions, the accelerator can perform polynomial multiplication in 2.46 ms and relinearization / key switching in 2.28 ms, showing poor flexibility in efficient execution. The remaining Karatsuba implementations continuously reduce computation time, improving algorithm efficiency.

[0075] Figure 2 A block diagram of a digital asset user analysis encryption system according to this application is shown.

[0076] like Figure 2 As shown, this application discloses a digital asset user analysis encryption system. The digital asset user analysis encryption system 2 includes a memory 21 and a processor 22. The memory 21 includes a digital asset user analysis encryption method program. When the digital asset user analysis encryption method program is executed by the processor 22, it performs the following steps:

[0077] Extract structured information from text documents containing behavioral and contextual data from social media and online platforms into comma-separated value (CSV) files;

[0078] The structured results processed by the text analysis platform are combined with other structured data sources, and the dimensional data is transformed and loaded into the underlying data warehouse using SNM's data integration tool (Zanho).

[0079] Integrated preprocessing of multi-source heterogeneous data in a data warehouse;

[0080] This system utilizes an artificial intelligence engine based on natural language processing / language engineering to analyze and predict pre-processed multi-source heterogeneous data, and outputs the analysis and prediction results.

[0081] It should be noted that in S102, a text analysis platform is used to extract structured information from text documents into comma-separated value (CSV) files. This structured result is then combined with other structured data sources. During the ETL process (transforming, cleaning, and loading the extracted text as a CSV file into the data warehouse), SNM's data integration tool (Zanho) loads it into the underlying data warehouse server. Fact data is transformed, mapped to dimensional data, and loaded into the data warehouse. The import of new data creates new tables for dimensions and facts. The import of incremental data updates or refreshes existing dimensional and fact tables.

[0082] It is worth mentioning the online multi-source heterogeneous data extraction and unstructured data warehouse design model for Social Networks Mining (SNM). This invention combines a text processing engine and a predefined extractor library to help identify and extract text documents containing behavioral and contextual data from multiple social media and online platforms, and further extract relevant structured items from them.

[0083] According to embodiments of the present invention, integrated preprocessing of multi-source heterogeneous data in a data warehouse includes:

[0084] Construct heterogeneous graph convolutional networks for multi-source heterogeneous data in data warehouses;

[0085] A novel two-layer attention mechanism is used to preprocess behavioral and scene data.

[0086] According to an embodiment of the present invention, the construction of the heterogeneous graph convolutional network specifically involves:

[0087] Define an undirected graph G=(V,E;O,S) that uses entity type mapping:φ:V→O and link type mapping:∅:E→S; vertex υ∈V is an entity, and edge e= = V,u>∈e represents the relationship between V and u, where υ,u∈V;

[0088] ti∈O={o1,o2,…,on} is an entity type, φ(υ)∈T;

[0089] sj∈S={s1,s2,…,sm} is a relation, ∅(e)∈R; the number of entity types |M|>1 and the number of relation types |R|>1; all vertices V={V1∪V2,…∪Vn} can be partitioned into n mutually exclusive subsets; all edges E={E1∪E2,…∪Em} can be partitioned into m mutually exclusive subsets;

[0090] In a weighted network, w(e) represents the weight of e = ⟨v, u>∈e.

[0091] According to an embodiment of the present invention, the novel two-layer attention mechanism includes a node-level attention machine and a type-level attention.

[0092] According to an embodiment of the present invention, the preprocessing of behavioral data and scene data using a novel dual-layer attention mechanism specifically involves: embedding HIN (Hetegeneous Information Network) into HGCN (Heterogeneous Graph Convolutional Network Deep Learning Model for Collective Classification), and then introducing a dual-layer attention mechanism to calculate the attention weights of different neighboring nodes and nodes of different types.

[0093] It should be noted that the novel two-layer attention mechanism based on heterogeneous graph convolutional networks is used for data analysis based on behavioral and scene data. It includes node-level attention and type-level attention. We embed HIN (Hetegeneous Information Network) into HGCN (Hypergeneous Graph Convolutional Network Deep Learning Model for Collective Classification), and then introduce the aforementioned two-layer attention mechanism to calculate the attention weights of different neighboring nodes and nodes of different types.

[0094]

[0095] This algorithm can solve the problems of semantic sparsity and hidden network structure information. It can also integrate several additional pieces of information, which can greatly enrich the semantic information.

[0096] According to an embodiment of the present invention, the analysis and prediction of preprocessed multi-source heterogeneous data using an artificial intelligence engine based on natural language processing / language engineering specifically involves:

[0097] First, collect several days' worth of data and store it in an Apache Spark RDD, then perform computations to build features from the raw data to guide the AI ​​engine;

[0098] Next, a data stream is initiated to obtain behavioral and contextual data and related real-time data from social media; features are calculated on this data to obtain new data points for future predictions from the artificial intelligence engine.

[0099] It should be noted that the natural language processing / language engineering-based engine described in this invention aims to analyze and predict using previously collected and preprocessed behavioral and contextual data. This process is achieved by learning the correlation between relevant behavioral and contextual data sentiment and the relevant digital asset users.

[0100] The engine first learns from a given history and then begins making predictions based on previous correlations. It is also able to improve itself through its mistakes and attempts to refine its predictions. Furthermore, the engine supports trend visualization of records over time based on incoming data and intermediate results. Built with the ever-increasing volume, speed, and diversity of available data in mind, the engine is capable of scaling and managing large amounts of heterogeneous data.

[0101] This engine is built within the Apache ecosystem and uses Apache Spark. The data structure in Spark is based on Resilient Distributed Datasets (RDDs), which are read-only, multi-set datasets that can be distributed across a group of machines and offer a degree of fault tolerance. Spark applications run as separate processes on different clusters and are coordinated by a Spark object (also known as the SparkContext). This element is the main driver of the program; it connects to the cluster manager and helps acquire executors on different nodes to allocate resources across the application. Spark is highly scalable, up to 100 times faster than Hadoop on large datasets, and provides out-of-the-box libraries for streaming and machine learning. At the heart of the architecture is Apache Spark, which acts as an in-memory data store and allows us to perform computations in a scalable manner. This data is the input to our machine learning model for making predictions. To bootstrap our model, we first collect several days' worth of data and store it in an Apache Spark RDD. Next, we perform computations to build features from the raw data. All of these computations are performed on data distributed across multiple Spark clusters and therefore scale as the data grows.

[0102] Once the AI ​​engine is bootstrapping, we begin a data stream to acquire behavioral and contextual data from social media and related real-time data. Similar computations are performed on this data to compute features, and this new data point is then used to derive future predictions from the model. This computed data point is then appended to the existing data in the Spark RDD, derived from the bootstrap data. Therefore, in addition to the ability to create user profiles, we are continuously expanding our data storage, enabling us to extract holistic visualizations from data on digital asset market trends.

[0103] A third aspect of the present invention provides a computer-readable storage medium comprising a digital asset user analysis encryption method program, wherein when executed by a processor, the digital asset user analysis encryption method program implements the steps of the digital asset user analysis encryption method as described in any of the above claims.

[0104] This invention provides crucial support for user information identification and selection, and precise profiling through the analysis of multi-source heterogeneous data targeting behavioral and contextual data, thereby improving customer satisfaction. By extracting behavioral and contextual information features such as emotional characteristics and textual feedback from individuals and groups online, this invention can identify core value customers. Using customer behavior data, feature data, and social interaction data as input dimensions, it can analyze customer preferences, as well as users' social influence and social networks, and further conduct research based on factors such as users' social status, economic conditions, and education level.

[0105] This invention proposes an online multi-source heterogeneous data extraction and unstructured data warehouse design model for Social Networks Mining (SNM). It helps identify and extract text documents containing behavioral and contextual data from multiple social media and online platforms, and further extracts relevant structured items. This improves identification efficiency while reducing information loss.

[0106] This invention proposes a novel two-layer attention mechanism based on heterogeneous graph convolutional networks for data analysis of behavioral and scene data. This mechanism can solve the problems of semantic sparsity and hidden network structure information, and can integrate additional information, thereby greatly enriching semantic information.

[0107] This invention proposes an engine based on natural language processing / language engineering that can expand and manage large amounts of heterogeneous data and visualize the trends of the results.

[0108] In summary, to balance privacy and security with operational efficiency, the core of the method proposed in this invention is a novel hardware algorithm for homomorphic encryption, which improves flexibility, scalability, and operating speed, thereby reducing computation time and improving algorithm efficiency.

[0109] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0110] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0111] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0112] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0113] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A digital asset user analysis encryption method, characterized in that, Includes the following steps: Using a text analytics platform, text analysis tools are used to extract structured information from text documents containing behavioral and contextual data from social media and online platforms into comma-separated value files; The structured results processed by the text analysis platform are combined with other structured data sources. The combined text is transformed and cleaned as a comma-separated value (CSV) file. The data integration tool of social network mining (SNM) then transforms the dimensional data and loads it into the underlying data warehouse server. Integrated preprocessing of multi-source heterogeneous data in a data warehouse includes: constructing a heterogeneous graph convolutional network for multi-source heterogeneous data in the data warehouse; and using a novel two-layer attention mechanism to preprocess behavioral data and scene data. This system utilizes an artificial intelligence engine based on natural language processing / language engineering to analyze and predict pre-processed multi-source heterogeneous data, and outputs the analysis and prediction results.

2. The encryption method for digital asset user analysis according to claim 1, characterized in that, The construction of the heterogeneous graph convolutional network specifically involves: Define an undirected graph This undirected graph uses entity type mapping: and link type mapping: ; vertex It is an entity, while the edge ,express and vertex The relationship between them, ; It is an entity type. ; It is a relationship. ; Number of entity types and the number of relation types ; All vertex sets , can be divided into n mutually exclusive subsets; Set of all edges , can be divided into m mutually exclusive subsets; In weighted networks, Representing an edge The weight value.

3. A digital asset user analysis encryption method according to claim 1 or 2, characterized in that, The novel two-layer attention mechanism includes a node-level attention mechanism and a type-level attention mechanism.

4. The encryption method for digital asset user analysis according to claim 3, characterized in that, The novel dual-layer attention mechanism is used to preprocess behavioral and scene data. Specifically, HIN is embedded into HGCN, and then a dual-layer attention mechanism is introduced to calculate the attention weights of different neighboring nodes and different types of nodes.

5. The encryption method for digital asset user analysis according to claim 4, characterized in that, The specific steps involve using an artificial intelligence engine based on natural language processing / language engineering to analyze and predict preprocessed multi-source heterogeneous data: First, collect several days' worth of data and store it in an Apache Spark RDD, then perform computations to build features from the raw data to guide the AI ​​engine; Next, a data stream is initiated to obtain real-time data related to behavioral and contextual data in social media; These data are used to calculate features, resulting in new data points used to make future predictions from an artificial intelligence engine.

6. A digital asset user analysis encryption system, characterized in that, The system includes a memory and a processor. The memory contains a digital asset user analysis encryption method program, which, when executed by the processor, performs the following steps: Using a text analytics platform, text analysis tools are used to extract structured information from text documents containing behavioral and contextual data from social media and online platforms into comma-separated value files; The structured results processed by the text analysis platform are combined with other structured data sources. The combined text is transformed and cleaned as a comma-separated value (CSV) file. The data integration tool of social network mining (SNM) then transforms the dimensional data and loads it into the underlying data warehouse server. Integrated preprocessing of multi-source heterogeneous data in a data warehouse includes: constructing a heterogeneous graph convolutional network for multi-source heterogeneous data in the data warehouse; and using a novel two-layer attention mechanism to preprocess behavioral data and scene data. This system utilizes an artificial intelligence engine based on natural language processing / language engineering to analyze and predict pre-processed multi-source heterogeneous data, and outputs the analysis and prediction results.

7. A digital asset user analysis encryption system according to claim 6, characterized in that, The specific steps involve using an artificial intelligence engine based on natural language processing / language engineering to analyze and predict preprocessed multi-source heterogeneous data: First, collect several days' worth of data and store it in an Apache Spark RDD, then perform computations to build features from the raw data to guide the AI ​​engine; Next, the data stream begins to obtain real-time data related to behavioral and contextual data in social media. These data are used to calculate features, resulting in new data points used to make future predictions from an artificial intelligence engine.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a digital asset user analysis encryption method program, which, when executed by a processor, implements the steps of a digital asset user analysis encryption method as described in any one of claims 1 to 5.

Images