Software development application data processing method based on cloud computing

Abstract

The invention discloses a software development application data processing method based on cloud computing, which can predict a computing demand and dynamically adjust the distribution of cloud computing resources according to a real-time load condition by introducing a dynamic resource distribution algorithm, thereby avoiding the defects of traditional fixed resource distribution, improving the resource utilization rate and improving the resource utilization rate. Through the improved K-Means algorithm and the random forest algorithm, the precision and efficiency of data clustering and classification are remarkably improved, especially when large-scale data are processed, the calculation complexity of the algorithm is optimized, the local optimum problem is avoided, and the method is suitable for large-scale data processing. And moreover, the calculation speed is improved through parallelization processing, the data transmission delay is effectively reduced by adopting a multi-stage cache mechanism, and particularly, the repeated calculation and transmission waiting time are reduced by dynamically adjusting a cache strategy between an edge node and a core node, so that the overall response speed of the system is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for processing software development applications based on cloud computing. Background Art

[0002] In existing cloud computing-based software development applications, the data processing process usually involves multiple links such as data collection, preprocessing, analysis and visualization. However, the existing technology has the following shortcomings:

[0003] 1. Low data processing efficiency: Traditional data processing methods rely on fixed computing resource allocation and cannot be dynamically adjusted according to real-time load conditions, which easily leads to resource waste or computing bottlenecks.

[0004] 2. Low algorithm accuracy: Existing clustering and classification algorithms have limited efficiency in large-scale data processing. In particular, the selection of the initial centroid has a significant impact on the results in the K-Means algorithm and is prone to falling into local optimality.

[0005] 3. Data transmission delay: There is a delay in the transmission of data between different nodes in the cloud computing platform, which affects the overall processing speed. Especially when processing large amounts of data, the transmission bottleneck is particularly obvious.

[0006] 4. Privacy and security issues: When processing user data, existing technologies lack effective encryption and privacy protection mechanisms, which can easily lead to the leakage of sensitive information.

[0007] 5. Data visualization is not intuitive enough: The presentation of high-dimensional data in existing visualization technologies is not intuitive enough, and it is difficult for users to conduct in-depth multi-dimensional analysis through interaction.

[0008] In response to the above problems, the present invention significantly improves the data processing efficiency, security and visualization effect on the cloud computing platform by introducing dynamic resource allocation, improved algorithms, intelligent caching mechanisms, encryption and privacy protection technologies. Summary of the invention

[0009] The purpose of the present invention is to provide a software development application data processing method based on cloud computing to solve the problems existing in the existing data processing methods proposed in the above background technology.

[0010] To achieve the above object, the present invention provides the following technical solution, a method for processing software development application data based on cloud computing, comprising the following steps:

[0011] S1. The data acquisition module obtains raw data from multiple heterogeneous data sources through the cloud computing platform. The data sources include structured data, unstructured data, real-time streaming data, and historical static data;

[0012] S2. The data preprocessing module performs data cleaning, format conversion, data deduplication, missing value filling and data standardization on the original data;

[0013] S3. Use dynamic resource allocation algorithms to allocate computing resources in a cloud computing environment and dynamically adjust computing resources based on the complexity and volume of data processing;

[0014] S4. Perform parallel data processing and analysis operations through cloud computing platforms;

[0015] S5. Provide real-time data visualization based on the processing results, and support dynamic update of data results and multi-dimensional interactive analysis.

[0016] Preferably, the dynamic resource allocation algorithm comprises the following steps:

[0017] S3.1. Use historical load data to establish a load forecasting model, and use the long short-term memory network (LSTM) to predict future data processing loads. The prediction accuracy is improved through multiple iterations of model training.

[0018] S3.2. Based on the predicted load results, use a linear programming algorithm to optimize resource allocation among different cloud computing nodes;

[0019] S3.3. Monitor the current load situation in real time, compare it with the predicted results, and use the feedback mechanism to adjust the resource allocation strategy.

[0020] Preferably, the objective function of the linear programming algorithm is:

[0021]

[0022] Among them, C i represents the usage cost of the ith cloud computing node, R i is the amount of resources allocated to the i-th node, n is the number of cloud computing nodes, D j represents the delay cost of data transmission, m is the number of nodes that need to be transmitted, and λ is the weight coefficient of the delay cost.

[0023] Preferably, the data processing and analysis operations include the following data analysis steps:

[0024] S4.1. The collected data are clustered using the improved K-Means algorithm. The improvement is to introduce a density peak detection algorithm to reduce the dependence on the initial centroid and improve the convergence speed and accuracy of the algorithm.

[0025] S4.2. Use the random forest algorithm to classify the clustered data. The algorithm performs parallel processing based on the high concurrent processing capability of the cloud computing platform to reduce classification time and improve classification accuracy.

[0026] Preferably, the improved K-Means algorithm comprises the following steps:

[0027] S4.1.1. Use the density peak detection algorithm to preliminarily determine the cluster center. The density peak detection is based on the local density of each data point and the distance calculation to other data points;

[0028] S4.1.2. Optimize the preliminary clustering results.

[0029] Preferably, the data processing adopts a multi-level cache mechanism to optimize the data processing speed, and the multi-level cache mechanism includes the following steps:

[0030] S4.1. Set up a first-level cache at the edge node of the cloud computing platform;

[0031] S4.2. Set up a secondary cache in the core nodes of cloud computing to store large-scale intermediate processing data and reduce repeated calculations;

[0032] S4.3. Use intelligent cache replacement strategy to dynamically adjust the data in the first and second level caches based on data access frequency and cache space.

[0033] Preferably, step S1 further includes a data encryption and privacy protection mechanism, and the data encryption and privacy protection mechanism includes the following steps:

[0034] S1.1. Use homomorphic encryption algorithms to encrypt sensitive data, ensuring that calculations can be completed without decryption during data processing, thereby improving data security;

[0035] S1.2. Differential privacy technology is used to add noise interference to the processing results to ensure that user privacy data will not be leaked during the analysis process.

[0036] Preferably, the encryption function of the homomorphic encryption algorithm is:

[0037] E(x)=g x mod p,

[0038] Among them, g is the generator, x is the data to be encrypted, and p is a large prime number.

[0039] Preferably, step S5 further includes a data visualization step, and the data visualization step includes:

[0040] S5.1. Use dimensionality reduction algorithms to map high-dimensional data to low-dimensional space, using PCA (principal component analysis) or t-SNE (t-distributed stochastic neighbor embedding) algorithms for dimensionality reduction;

[0041] S5.2. Visual charts are generated through the cloud-based real-time rendering engine, including scatter plots, line charts and heat maps. Users can conduct in-depth analysis of data based on multi-dimensional interactive operations.

[0042] Preferably, step S5 further includes a relationship analysis step based on a graph neural network for analyzing the correlation between different data nodes, and the relationship analysis step includes:

[0043] S5.3. Build a graph structure model of data, where data nodes are graph vertices and the relationships between data are edges;

[0044] S5.4. Use graph neural network algorithms to learn and analyze data nodes and mine complex relationships between data.

[0045] Compared with the prior art, the cloud computing-based software development application data processing method of the present invention has the following beneficial effects:

[0046] 1. Dynamic resource allocation improves efficiency: By introducing a dynamic resource allocation algorithm, the present invention can predict computing needs according to real-time load conditions and dynamically adjust the allocation of cloud computing resources, thereby avoiding the shortcomings of traditional fixed resource allocation, improving resource utilization, and significantly improving system performance and stability under high concurrency conditions.

[0047] 2. Algorithm optimization and enhanced data processing: Through the improved K-Means algorithm and random forest algorithm, the present invention significantly improves the accuracy and efficiency of data clustering and classification, especially when processing large-scale data, optimizes the computational complexity of the algorithm, avoids local optimal problems, and improves the calculation speed through parallel processing.

[0048] 3. Multi-level cache reduces latency: The multi-level cache mechanism effectively reduces data transmission latency, especially by dynamically adjusting the cache strategy between edge nodes and core nodes, reducing repeated calculations and transmission waiting time, and further improving the overall response speed of the system.

[0049] 4. Enhanced security and privacy protection: Through homomorphic encryption and differential privacy technology, the present invention ensures the security of data during processing. Calculations can be completed even when the data is encrypted, protecting the privacy of sensitive data and avoiding the risk of data leakage. It is suitable for application scenarios that require high security.

[0050] 5. Efficient data visualization and in-depth analysis: The present invention uses dimensionality reduction technology and graph neural networks for data visualization and association analysis, which can intuitively present high-dimensional data to users, and through in-depth analysis of the complex relationships between data, help users more comprehensively understand the potential value of the data, thereby enhancing the depth and accuracy of data analysis.

[0051] 6. Adaptability to multiple data types and scenarios: The present invention supports unified processing of multiple types of data (including structured data, unstructured data, real-time data and historical data), has strong flexibility and scalability, and can be widely used in different data analysis and processing scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 The present invention is a flowchart of a method for processing software development application data based on cloud computing. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In the description of the invention, it should be noted that the terms "upper", "lower", "inner", "outer", "front end", "rear end", "two ends", "one end", "the other end" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0055] In the description of the invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "provided with", "connected", etc. should be understood in a broad sense. For example, "connected" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0056] See also Figure 1 The present invention provides a technical solution: a method for processing software development application data based on cloud computing, comprising the following steps:

[0057] S1. The data acquisition module obtains raw data from multiple heterogeneous data sources through the cloud computing platform. The data sources include structured data, unstructured data, real-time streaming data, and historical static data;

[0058] S2. The data preprocessing module performs data cleaning, format conversion, data deduplication, missing value filling and data standardization on the raw data to ensure the integrity and consistency of the data;

[0059] S3. Use dynamic resource allocation algorithms to allocate computing resources in a cloud computing environment, dynamically adjust computing resources based on the complexity and volume of data processing, and ensure system stability under high concurrency conditions;

[0060] S4. Perform parallel data processing and analysis operations through cloud computing platforms to improve data processing efficiency;

[0061] S5. Provide real-time data visualization based on the processing results, and support dynamic update of data results and multi-dimensional interactive analysis.

[0062] Specifically, through a complete data processing flow, the quality of data and the efficiency of processing are ensured, and different data sources and processing requirements are adapted. Dynamic resource allocation and parallel processing improve the processing speed and system stability, especially when facing large-scale data.

[0063] The dynamic resource allocation algorithm includes the following steps:

[0064] S3.1. Use historical load data to establish a load forecasting model, and use the long short-term memory network (LSTM) to predict future data processing loads. The prediction accuracy is improved through multiple iterations of model training.

[0065] S3.2. Based on the predicted load results, use the linear programming algorithm to optimize resource allocation between different cloud computing nodes to ensure efficient use of computing resources;

[0066] S3.3. Monitor the current load situation in real time, compare it with the predicted results, and use the feedback mechanism to adjust the resource allocation strategy to avoid resource overload or waste.

[0067] Specifically, through load prediction and optimization algorithms, resource allocation is adjusted dynamically, resource waste is reduced, computing efficiency is improved, and real-time adjustment strategies are used to reduce resource overload or shortage, ensuring stable operation of the system under high load conditions.

[0068] The objective function of the linear programming algorithm is:

[0069]

[0070] Among them, C i represents the usage cost of the ith cloud computing node, R i is the amount of resources allocated to the i-th node, n is the number of cloud computing nodes, D jrepresents the delay cost of data transmission, m is the number of nodes that need to be transmitted, and λ is the weight coefficient of the delay cost.

[0071] Specifically, the linear programming algorithm used in dynamic resource allocation considers not only the computational cost but also the data transmission delay cost in the objective function, comprehensively considering the computational and transmission costs to ensure comprehensive optimization of resource allocation and improve the performance of the overall system.

[0072] The data processing and analysis operation includes the following data analysis steps:

[0073] S4.1. The collected data are clustered using the improved K-Means algorithm. The improvement is to introduce a density peak detection algorithm to reduce the dependence on the initial centroid and improve the convergence speed and accuracy of the algorithm.

[0074] S4.2. Use the random forest algorithm to classify the clustered data. The algorithm performs parallel processing based on the high concurrent processing capability of the cloud computing platform to reduce classification time and improve classification accuracy.

[0075] Specifically, the improved K-Means algorithm and random forest algorithm improve the accuracy and efficiency of data processing, especially in big data environments.

[0076] The improved K-Means algorithm includes the following steps:

[0077] S4.1.1. Use the density peak detection algorithm to preliminarily determine the cluster center. The density peak detection is based on the local density of each data point and the distance calculation to other data points;

[0078] S4.1.2. Optimize the preliminary clustering results, optimize the location of the cluster center through local search, reduce local optimal problems, and ensure the global optimality of the clustering results.

[0079] Specifically, the improved K-Means algorithm, including density peak detection and local optimization, reduces the impact of initial centroid selection on the results and improves clustering accuracy and convergence speed.

[0080] Data processing uses a multi-level cache mechanism to optimize data processing speed. The multi-level cache mechanism includes the following steps:

[0081] S4.1. Set up a first-level cache at the edge node of the cloud computing platform to store recently accessed high-frequency data and reduce data transmission delay;

[0082] S4.2. Set up a secondary cache in the core nodes of cloud computing to store large-scale intermediate processing data and reduce repeated calculations;

[0083] S4.3. Use intelligent cache replacement strategy to dynamically adjust the data in the first and second level caches according to data access frequency and cache space to improve data processing speed.

[0084] Specifically, the intelligent caching mechanism reduces the delay in data transmission and processing, and improves the response speed of the overall system.

[0085] Among them, step S1 also includes a data encryption and privacy protection mechanism, which includes the following steps:

[0086] S1.1. Use homomorphic encryption algorithms to encrypt sensitive data, ensuring that calculations can be completed without decryption during data processing, thereby improving data security;

[0087] S1.2. Differential privacy technology is used to add noise interference to the processing results to ensure that user privacy data will not be leaked during the analysis process.

[0088] Specifically, homomorphic encryption ensures the security of data processing, and differential privacy technology protects user privacy data, and is suitable for scenarios that process sensitive information.

[0089] The encryption function of the homomorphic encryption algorithm is:

[0090] E(x)=g x mod p,

[0091] Among them, g is the generator, x is the data to be encrypted, and p is a large prime number, which ensures that sensitive data can be calculated without decryption during the processing process, thereby improving data security and processing efficiency.

[0092] Specifically, the security of data is protected by encryption functions, and effective calculations can be performed even when the data is in an encrypted state.

[0093] Among them, step S5 also includes a data visualization step, and the data visualization step includes:

[0094] S5.1. Use dimensionality reduction algorithms to map high-dimensional data to low-dimensional space, using PCA (principal component analysis) or t-SNE (t-distributed stochastic neighbor embedding) algorithms for dimensionality reduction;

[0095] S5.2. Visual charts are generated through the cloud-based real-time rendering engine, including scatter plots, line charts and heat maps. Users can conduct in-depth analysis of data based on multi-dimensional interactive operations.

[0096] Specifically, the visualization of data is improved through dimensionality reduction and real-time rendering, and users can understand and analyze data more intuitively.

[0097] Among them, step S5 also includes a relationship analysis step based on a graph neural network, which is used to analyze the correlation between different data nodes. The relationship analysis step includes:

[0098] S5.3. Build a graph structure model of data, where data nodes are graph vertices and the relationships between data are edges;

[0099] S5.4. Use graph neural network algorithms to learn and analyze data nodes, explore complex relationships between data, and improve the accuracy of analysis.

[0100] Specifically, by analyzing the complex relationships between data through graph neural networks, the ability to understand and analyze data correlations is improved.

[0101] Example 1: Dynamic resource allocation and computing optimization

[0102] Background: Real-time streaming data is processed on a cloud computing platform. The data includes 100TB of structured data and 50TB of unstructured data, and the processing process requires dynamic adjustment of resource allocation.

[0103] Steps and data:

[0104] 1. Load Forecasting:

[0105] Historical data shows that processing 10TB of data per hour requires about 1,000 computing nodes.

[0106] The LSTM model is used to predict that the data processing load in the next hour will be 12TB.

[0107] 2. Linear programming objective function:

[0108] Compute the usage cost of the node C i The price is 0.1 yuan / hour / node.

[0109] Data transmission delay cost D j It is 0.05 yuan / GB.

[0110] Assume R i The amount of resources allocated to each node, the objective function is:

[0111]

[0112] Where n = 1000, m = 12TB × 1024MB / GB = 12288GB, λ = 0.05,

[0113] calculate:

[0114]

[0115] Assuming uniform resource distribution,

[0116] R i =12TB / 1000=12GB,

[0117] Calculation cost = 1000 × 0.1 × 1.2 = 1200,

[0118] Transmission delay cost = 0.05 × 12288 = 614.4,

[0119] Z=1200+614.4=1814.4.

[0120] Effect: By optimizing resource allocation and considering transmission delay costs, cost-effectiveness is maximized and processing stability is ensured under high load conditions.

[0121] Example 2: Data clustering and classification

[0122] Background: A dataset containing 1 million user behavior records is processed. The dataset includes user click records, purchase records, and search records.

[0123] Steps and data:

[0124] 1. Data preprocessing:

[0125] First, the user behavior records are cleaned, duplicate records are removed, and missing values ​​are filled. After preprocessing, the data set contains 500,000 valid records.

[0126] Normalize the data and scale the data to the [0,1] interval.

[0127] 2. Improved K-Means algorithm:

[0128] The density peak detection algorithm is used to determine the preliminary cluster centers. The algorithm is calculated based on the local density and distance of the data points to select appropriate preliminary cluster centers. Assume that 5 centers are initially selected.

[0129] The initially selected cluster centers are locally optimized and their positions are updated to reduce the distance from each data point to its nearest cluster center, thereby improving the accuracy and effect of clustering.

[0130] Calculate cluster centers:

[0131] The cluster centers initially selected are C1, C2, C3, C4, and C5, which represent different user behavior patterns.

[0132] 3. Random Forest Classification:

[0133] Use the processed data set to train a random forest classification model. Set the number of decision trees in the random forest to 100 and the maximum depth of the tree to 10.

[0134] The trained model is evaluated using a test set (e.g. 20% of the data randomly extracted from the dataset). The test set is data that was not used in training and is used to verify the generalization ability of the model.

[0135] Accuracy calculation:

[0136] 1. Model prediction:

[0137] The data in the test set is input into the trained random forest model to obtain the predicted label for each test sample.

[0138] 2. Accuracy calculation:

[0139] True label: The true class label of each data point in the test set.

[0140] Predicted Label: The model’s predicted class label for each data point in the test set.

[0141] Accuracy formula:

[0142]

[0143] Calculation: Assuming that out of the 20,000 samples in the test set, the model correctly predicted the category of 17,000 samples, then:

[0144]

[0145] Model performance evaluation: After the above process, an accuracy of 85% was obtained. This shows that the random forest model performs well in the classification task and can correctly classify 85% of the test data.

[0146] Effectiveness verification: The accuracy results show that the improved K-Means algorithm and random forest classifier effectively process user behavior data, can accurately classify the data, and provide valuable classification information.

[0147] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for processing software development application data based on cloud computing, characterized in that: The following steps are involved: S1. The data acquisition module obtains raw data from multiple heterogeneous data sources through the cloud computing platform. The data sources include structured data, unstructured data, real-time streaming data, and historical static data; S2. The data preprocessing module performs data cleaning, format conversion, data deduplication, missing value filling and data standardization on the original data; S3. Use dynamic resource allocation algorithms to allocate computing resources in a cloud computing environment and dynamically adjust computing resources based on the complexity and volume of data processing; S4. Perform parallel data processing and analysis operations through cloud computing platforms; S5. Provide real-time data visualization based on the processing results, and support dynamic update of data results and multi-dimensional interactive analysis.

2. The method for processing software development and application data based on cloud computing according to claim 1, characterized in that , The dynamic resource allocation algorithm comprises the following steps: S3.

1. Use historical load data to establish a load forecasting model, and use the long short-term memory network (LSTM) to predict future data processing loads. The prediction accuracy is improved through multiple iterations of model training. S3.

2. Based on the predicted load results, use a linear programming algorithm to optimize resource allocation among different cloud computing nodes; S3.

3. Monitor the current load situation in real time, compare it with the predicted results, and use the feedback mechanism to adjust the resource allocation strategy.

3. The method for processing software development and application data based on cloud computing according to claim 2, characterized in that , The objective function of the linear programming algorithm is: Among them, C i represents the usage cost of the ith cloud computing node, R i is the amount of resources allocated to the i-th node, n is the number of cloud computing nodes, D j represents the delay cost of data transmission, m is the number of nodes that need to be transmitted, and λ is the weight coefficient of the delay cost.

4. The method for processing software development and application data based on cloud computing according to claim 1, characterized in that: The data processing and analysis operations include the following data analysis steps: S4.

1. The collected data are clustered using the improved K-Means algorithm. The improvement is to introduce a density peak detection algorithm to reduce the dependence on the initial centroid and improve the convergence speed and accuracy of the algorithm. S4.

2. Use the random forest algorithm to classify the clustered data. The algorithm performs parallel processing based on the high concurrent processing capability of the cloud computing platform to reduce classification time and improve classification accuracy.

5. The method for processing software development and application data based on cloud computing according to claim 4, characterized in that: The improved K-Means algorithm comprises the following steps: S4.1.

1. Use the density peak detection algorithm to preliminarily determine the cluster center. The density peak detection is based on the local density of each data point and the distance calculation to other data points; S4.1.

2. Optimize the preliminary clustering results.

6. The method for processing software development and application data based on cloud computing according to claim 1, characterized in that: The data processing adopts a multi-level cache mechanism to optimize the data processing speed, and the multi-level cache mechanism includes the following steps: S4.

1. Set up a first-level cache at the edge node of the cloud computing platform; S4.

2. Set up a secondary cache in the core nodes of cloud computing to store large-scale intermediate processing data and reduce repeated calculations; S4.

3. Use intelligent cache replacement strategy to dynamically adjust the data in the first and second level caches based on data access frequency and cache space.

7. The method for processing software development and application data based on cloud computing according to claim 1, It is characterized in that Among them, step S1 also includes a data encryption and privacy protection mechanism, and the data encryption and privacy protection mechanism includes the following steps: S1.

1. Use homomorphic encryption algorithms to encrypt sensitive data, ensuring that calculations can be completed without decryption during data processing, thereby improving data security; S1.

2. Differential privacy technology is used to add noise interference to the processing results to ensure that user privacy data will not be leaked during the analysis process.

8. The method for processing software development and application data based on cloud computing according to claim 7, characterized in that: The encryption function of the homomorphic encryption algorithm is: E(x)=g x mod p, Among them, g is the generator, x is the data to be encrypted, and p is a large prime number.

9. The method for processing software development and application data based on cloud computing according to claim 1, characterized in that: in, Step S5 also includes a data visualization step, and the data visualization step includes: S5.

1. Use dimensionality reduction algorithms to map high-dimensional data to low-dimensional space, using PCA (principal component analysis) or t-SNE (t-distributed stochastic neighbor embedding) algorithms for dimensionality reduction; S5.

2. Visual charts are generated through the cloud-based real-time rendering engine, including scatter plots, line charts and heat maps. Users can conduct in-depth analysis of data based on multi-dimensional interactive operations.

10. The method for processing software development and application data based on cloud computing according to claim 1, It is characterized in that Among them, step S5 also includes a relationship analysis step based on a graph neural network, which is used to analyze the correlation between different data nodes. The relationship analysis step includes: S5.

3. Build a graph structure model of data, where data nodes are graph vertices and the relationships between data are edges; S5.

4. Use graph neural network algorithms to learn and analyze data nodes and mine complex relationships between data.

Images