Data mining method and system based on industrial park platform

Abstract

The present invention discloses a data mining method and system based on an industrial park platform, including obtaining raw data from the industrial park platform and preprocessing it, generating new output data from the preprocessed data through a neural network; based on the generated new output data, analyzing historical data points through an autoregressive model to predict future values, and then analyzing and predicting the dynamic changes of data in different dimensions through a diffusion model; establishing a mapping relationship between a preset knowledge graph and an industrial park platform database, matching the output data with knowledge items in a preset knowledge graph, and determining knowledge items that match the output data; the present invention significantly improves the analysis accuracy of time series data and multi-dimensional data by introducing neural networks, autoregressive models and diffusion models, and utilizes the deep integration of knowledge graphs and industrial park platform data to enhance the practicability and interpretability of data analysis and realize intelligent data mining.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and in particular to a data mining method and system based on an industrial park platform. Background Art

[0002] The existing data mining technologies in the application of industrial park platforms mainly focus on using basic statistical analysis and machine learning algorithms to process and analyze data. These technologies include simple decision trees, as well as some traditional clustering and classification algorithms. Although these methods can provide useful insights in some cases, they usually do not take into account the temporal characteristics of the data, the multi-dimensional diffusion characteristics, and the deep integration of data and domain knowledge, resulting in inaccuracy in the analysis of industrial data with complex temporal associations and multi-dimensional characteristics. Summary of the invention

[0003] The purpose of the present invention is to provide a data mining method and system based on an industrial park platform, which improves the analysis accuracy of time series data and multidimensional data, and enhances the ability to deeply integrate data analysis with domain knowledge.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] This application provides a data mining method based on an industrial park platform, comprising the following steps:

[0006] S1, obtain the original data from the industrial park platform and preprocess it, and generate new output data through the neural network;

[0007] Among them, the neural network is a neural network with multiple layers and nested loop layers, and the back propagation algorithm is used to train the neural network and optimize the model parameters;

[0008] S2. Based on the generated new output data, an autoregressive model and a diffusion model are established. The time series data is predicted by the autoregressive model, and the historical data points are analyzed to predict future values. The diffusion model is then used to simulate the diffusion of data in different dimensions, and the dynamic changes of data in different dimensions are analyzed and predicted.

[0009] S3. After obtaining new output data generated by the neural network, a mapping relationship between a preset knowledge graph and an industrial park platform database is established, the output data is matched with knowledge items in the preset knowledge graph, and knowledge items matching the output data are determined;

[0010] Among them, the preset knowledge graph includes the domain knowledge and information structure of the industrial park platform.

[0011] Furthermore, before obtaining the original data from the industrial park platform, it also includes: obtaining target data; the target data determines the data of the industrial park platform to be mined.

[0012] Furthermore, a neural network is a neural network with multiple layers and nested recurrent layers, specifically including:

[0013] The neural network includes a plurality of layers, each layer includes a set of nodes, and the neural network also includes a nested loop layer, the nested loop layer includes a first loop layer and a second loop layer, the first loop layer is configured to send a message to the second loop layer.

[0014] Furthermore, the back propagation algorithm is used to train the neural network and optimize the model parameters, including:

[0015] The preprocessed data is input into the neural network, and the weights and activation functions are transformed through the nodes of each layer to obtain new output data;

[0016] The loss function is used to calculate the gap between the new output data of the neural network and the target value, and the loss value of the current model is obtained;

[0017] According to the loss value, the gradient of each weight data of the loss function is calculated by the chain rule; wherein the gradient represents the rate of change of the loss function relative to the weight data; according to the calculated gradient, the weight data in the neural network is adjusted and updated;

[0018] By iteratively training the neural network, the weight data is continuously updated so that the model output data fits the training data.

[0019] Furthermore, we use autoregressive models to predict time series data and analyze historical data points to predict future values. The specific contents include:

[0020] Determine the order of the autoregressive model and identify autocorrelation in the data by analyzing the autocorrelation function and partial autocorrelation function of time series data;

[0021] Use the least squares method to estimate the parameters of the autoregressive model, including the autoregressive coefficients and the constant term;

[0022] Apply the estimated parameters to the autoregressive model to fit historical data, and then perform model diagnosis;

[0023] Use the fitted autoregressive model to predict future observations and predict the observations in h steps;

[0024] Specifically, the value of h steps is predicted by a recursive formula, which is expressed as:

[0025]

[0026] in, represents the observed value predicted h steps later at time point t, c represents the model parameters, is the calculated forecast value, which is the best estimate made at time point t. is the observed value of hp predicted at time point t, and p is the lag order, which is used to define the number of historical observations that the autoregressive model uses when making predictions.

[0027] Furthermore, the diffusion model is used to simulate the diffusion of data in different dimensions, analyze and predict the dynamic changes of data in different dimensions. Specifically, the diffusion process of data is defined by stochastic differential equations, and the dynamic changes of data in different dimensions due to drift and diffusion effects are captured and predicted by solving the stochastic differential equations.

[0028] The specific stochastic differential equation is expressed as:

[0029] in, is a multidimensional state vector, f is a drift function, is the diffusion term matrix, is a multidimensional Brownian motion vector.

[0030] Furthermore, a mapping relationship between a preset knowledge graph and an industrial park platform database is established, the output data is matched with knowledge items in the preset knowledge graph, and knowledge items matching the output data are determined, including: converting the output data into a query request according to the mapping relationship; sending a query request to the industrial park platform database, and receiving the query result returned by the industrial park platform database; and using the query result as a knowledge item matching the output data.

[0031] Furthermore, before matching the output data with knowledge items in a preset knowledge graph, it also includes: obtaining the semantic relationship between each knowledge item in the preset knowledge graph; matching the output data with the knowledge items in the preset knowledge graph.

[0032] Furthermore, after matching the output data with the knowledge items in the preset knowledge graph, it also includes: receiving a corpus associated with multiple queries from multiple users from the industrial park platform database; wherein the query-associated corpus includes query requests and query results, and inputting the corpus into the neural network to generate an output, wherein the neural network is configured to: determine whether a word in the corpus is a stop word; when the word is not a stop word, assign a weight value to the word, and determine the importance of the word based on the assigned weight value; use a machine learning algorithm to adjust the hyperparameters of the neural network.

[0033] The data mining system based on the industrial park platform includes data preprocessing and feature extraction modules, model building and analysis modules, and knowledge graph integration and intelligent decision support modules.

[0034] Data preprocessing and feature extraction module, used to extract features from the raw data collected from the industrial park platform;

[0035] Model building and analysis module, integrating autoregressive models to predict time series data, introducing diffusion models to simulate the diffusion process of data in different dimensions, training neural networks with multiple layers and nested loop layers, optimizing model parameters through back propagation algorithms, and performing deep learning and pattern recognition of data;

[0036] The knowledge graph integration and intelligent decision support module builds and maintains the knowledge graph, including the domain knowledge and information structure of the industrial park platform, establishes the mapping relationship between the knowledge graph and the industrial park platform database, and effectively matches data and knowledge items;

[0037] It uses semantic relationships and query processing technology to convert output data into query requests, send them to the database, and process the returned results.

[0038] The beneficial effects of the present invention are:

[0039] (1) By introducing neural networks and autoregressive models, the ability to analyze the time series characteristics of data is improved. The present invention uses a neural network with multiple layers and nested loop layers to train the preprocessed data, optimizes the model parameters through the back propagation algorithm, and then uses the autoregressive model to predict time series data, analyze historical data points and predict future values, which can more accurately capture the time dependency of data and improve the analysis accuracy of time series data;

[0040] (2) This paper improves the ability to analyze multi-dimensional diffusion characteristics by introducing a diffusion model to simulate the diffusion of data in different dimensions. The diffusion model uses stochastic differential equations to define the diffusion process of data and captures and predicts the dynamic changes of data in different dimensions due to drift and diffusion effects by solving stochastic differential equations. It can more accurately reflect the diffusion characteristics of data in different dimensions, thereby improving the accuracy of multi-dimensional data analysis.

[0041] (3) By establishing a mapping relationship between a preset knowledge graph and the industrial park platform database, the problem of insufficient deep integration of data and domain knowledge is solved. After the neural network generates new output data, the method of the present invention matches the output data with the knowledge items in the preset knowledge graph and determines the knowledge items that match the output data, thereby ensuring that the data analysis results can be deeply integrated with the domain knowledge and improving the practicality and interpretability of data analysis. By obtaining the semantic relationship in the knowledge graph and converting the query request into an operational data query process, the relevance and matching accuracy of data analysis and domain knowledge are further enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0043] Figure 1 A schematic diagram of the process of the data mining method based on the industrial park platform provided in Example 1 of the present application;

[0044] Figure 2 A schematic diagram of a process for training a neural network using a back propagation algorithm using a data mining method based on an industrial park platform provided in Example 1 of the present application;

[0045] Figure 3 A schematic diagram of the process of predicting time series data using an autoregressive model in a data mining method based on an industrial park platform provided in Example 1 of the present application. DETAILED DESCRIPTION

[0046] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are only examples of methods and systems consistent with some aspects of the present application as detailed in the attached claims.

[0047] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0048] The specific implementation methods, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0049] Example 1

[0050] See also Figure 1-Figure 3 This embodiment provides a data mining method and system based on the industrial park platform, which improves the analysis accuracy of time series data and multi-dimensional data, and enhances the ability to deeply integrate data analysis with domain knowledge.

[0051] The present invention provides a data mining method based on an industrial park platform, comprising the following steps:

[0052] S1. Obtain raw data from the industrial park platform and preprocess it. The preprocessing steps include data cleaning, standardization and feature extraction. The preprocessed data is used to generate new output data through a neural network.

[0053] Among them, the neural network is a neural network with multiple layers and nested loop layers, and the back propagation algorithm is used to train the neural network and optimize the model parameters;

[0054] S2. Based on the generated new output data, an autoregressive model and a diffusion model are established. The time series data is predicted by the autoregressive model, and the historical data points are analyzed to predict future values. The diffusion model is then used to simulate the diffusion of data in different dimensions, and the dynamic changes of data in different dimensions are analyzed and predicted.

[0055] S3. After obtaining new output data generated by the neural network, a mapping relationship between a preset knowledge graph and an industrial park platform database is established, the output data is matched with knowledge items in the preset knowledge graph, and knowledge items matching the output data are determined;

[0056] Among them, the preset knowledge graph includes the domain knowledge and information structure of the industrial park platform.

[0057] Data mining based on the industrial park platform can effectively mine and analyze the data of the industrial park platform, not only generating valuable new data, but also associating these data with existing knowledge, thus providing a more powerful basis for decision support and intelligent analysis.

[0058] Furthermore, before obtaining the original data from the industrial park platform, it also includes: obtaining target data; the target data determines the data of the industrial park platform to be mined.

[0059] Furthermore, a neural network is a neural network with multiple layers and nested recurrent layers, specifically including:

[0060] The neural network includes a plurality of layers, each layer includes a set of nodes, and the neural network also includes a nested loop layer, the nested loop layer includes a first loop layer and a second loop layer, the first loop layer is configured to send a message to the second loop layer.

[0061] Specifically, in a neural network with nested loop layers, the first loop layer and the second loop layer work together to process sequence data and capture dynamic characteristics in time series. This layer is responsible for receiving the initial data points of the input sequence and performing preliminary processing on them. It usually contains a set of nodes that can apply weights and activation functions to the input data to extract features and transform data; the second loop layer receives information from the first loop layer and further processes this information. The node set of this layer can perform more complex feature extraction and data transformation, while taking into account the state information received from the first layer. The design of the second loop layer allows the network to transfer and update states between different time steps. The information transfer between the two loop layers is achieved through the recursive structure of the neural network. At each time step, the first loop layer processes the input data and generates output, and then uses this output as one of the inputs of the second loop layer, imitating the recursive structure in natural language processing, so that the network can process data with a hierarchical structure.

[0062] Furthermore, the back propagation algorithm is used to train the neural network and optimize the model parameters, including:

[0063] S11, input the preprocessed data into the neural network, and calculate through the nodes of each layer. In each layer, the input data is weighted and transformed by the activation function to obtain new output data;

[0064] S12, using the loss function to calculate the gap between the new output data of the neural network and the target value, and obtain the loss value of the current model;

[0065] S13. Calculate the gradient of each weight data of the loss function by the chain rule according to the loss value; wherein the gradient represents the rate of change of the loss function relative to the weight data; and adjust and update the weight data in the neural network according to the calculated gradient;

[0066] The specific update weight formula is: ,in, is the updated weight, is the weight before updating, is the learning rate, is the gradient of the loss function with respect to the weights.

[0067] S14. By iteratively training the neural network, the weight data is continuously updated so that the model output data fits the training data.

[0068] Specifically, during the training process, the hyperparameters are adjusted according to the performance of the model. During the hyperparameter adjustment process, the learning rate must be adjusted first. This is a key parameter that affects the training speed and stability. The learning rate can be gradually reduced during the training process through learning rate decay, or the learning rate can be dynamically adjusted according to the performance of the validation set. If the learning rate is set too high, the model may become unstable and difficult to converge; if it is too low, the convergence speed will be slower. Secondly, it is also important to adjust the batch size: a larger batch size helps the stability of training but requires more resources, while a smaller batch size can speed up weight updates. Finally, increasing the number of layers and nodes in the network can enhance the expressiveness of the model, but a balance needs to be found between complexity and performance to avoid overfitting.

[0069] Furthermore, we use autoregressive models to predict time series data and analyze historical data points to predict future values. The specific contents include:

[0070] S21. Determine the order of the autoregressive model and identify the autocorrelation in the data by analyzing the autocorrelation function and partial autocorrelation function of the time series data;

[0071] S22. Use the least squares method to estimate the parameters of the autoregressive model, including the autoregressive coefficients and the constant term;

[0072] S23, applying the estimated parameters to the autoregressive model to fit the historical data, and then performing model diagnosis to check whether the residual term appears as white noise, ensuring that the model has no omitted variables or overfitting;

[0073] S24, using the fitted autoregressive model to predict future observations, predicting the observations in h steps;

[0074] Specifically, the value of h steps is predicted by a recursive formula, which is expressed as:

[0075]

[0076] in, represents the observed value predicted h steps later at time point t, c represents the model parameters, is the calculated forecast value, which is the best estimate made at time point t. is the observed value of hp predicted at time point t, and p is the lag order, which is used to define the number of historical observations used by the autoregressive model when making current value predictions.

[0077] Each forecast step depends on the previous forecast and actual value, allowing the model to gradually adjust its forecasts for the future to more accurately reflect the dynamic changes in the time series.

[0078] Furthermore, the diffusion model is used to simulate the diffusion of data in different dimensions, analyze and predict the dynamic changes of data in different dimensions. Specifically, the diffusion process of data is defined by stochastic differential equations, and the dynamic changes of data in different dimensions due to drift and diffusion effects are captured and predicted by solving the stochastic differential equations.

[0079] The specific stochastic differential equation is expressed as:

[0080] in, is a multidimensional state vector, f is a drift function, is the diffusion term matrix, is a multidimensional Brownian motion vector.

[0081] Furthermore, a mapping relationship between a preset knowledge graph and an industrial park platform database is established, the output data is matched with knowledge items in the preset knowledge graph, and knowledge items matching the output data are determined, including: converting the output data into a query request according to the mapping relationship; sending a query request to the industrial park platform database, and receiving the query result returned by the industrial park platform database; and using the query result as a knowledge item matching the output data.

[0082] Furthermore, before matching the output data with knowledge items in a preset knowledge graph, it also includes: obtaining the semantic relationship between each knowledge item in the preset knowledge graph; matching the output data with the knowledge items in the preset knowledge graph.

[0083] Furthermore, after matching the output data with the knowledge items in the preset knowledge graph, it also includes: receiving a corpus associated with multiple queries from multiple users from the industrial park platform database; wherein the query-associated corpus includes query requests and query results, and inputting the corpus into the neural network to generate an output, wherein the neural network is configured to: determine whether a word in the corpus is a stop word; when the word is not a stop word, assign a weight value to the word, and determine the importance of the word based on the assigned weight value; use a machine learning algorithm to adjust the hyperparameters of the neural network so that the performance of the neural network is improved.

[0084] The data mining method based on the industrial park platform is a comprehensive and multi-level analysis process. The whole process starts with obtaining the raw data of the industrial park platform. After thorough preprocessing, including data cleaning, standardization and feature extraction, the data is input into a complex neural network, which consists of multiple nested loop layers to capture the complex patterns and relationships in the data. Through the back-propagation algorithm, the parameters of the model are optimized to minimize the prediction error.

[0085] After generating new output data, the autoregressive model is further used to predict time series data, using current and past observations to infer future trends. At the same time, the diffusion model is introduced to simulate the diffusion of data in different dimensions, which not only increases the understanding of the dynamic changes of data, but also improves the ability to predict data changes in many aspects. In order to combine these data with the existing knowledge system, a preset knowledge graph is established, which contains the domain knowledge and information structure of the industrial park platform. Through precise mapping relationships, the output data is matched with the knowledge items in the knowledge graph, which not only deepens the understanding of the data, but also provides strong support for decision-making.

[0086] In addition, corpora related to user queries are received from the industrial park platform database and input into the neural network to generate more accurate outputs. In this process, the neural network is configured to recognize and process non-stop words and assign weights to these words to determine their importance. Through machine learning algorithms, we continuously adjust the hyperparameters of the neural network to improve the performance of the model. During the entire training process, the adjustment of hyperparameters is crucial. The learning rate, as one of the key parameters, directly affects the speed and stability of model training. The learning rate is decayed or dynamically adjusted according to the performance of the validation set to avoid model instability or slow convergence. At the same time, the adjustment of batch size also has an important impact on training stability and resource requirements. Increasing the number of layers and nodes in the network can improve the expressiveness of the model, but it is also necessary to find a balance between complexity and performance to prevent overfitting. Through this series of comprehensive data mining steps, not only can valuable new data be generated, but these data can also be deeply associated with existing knowledge, providing a solid foundation for decision support and intelligent analysis of the industrial park platform.

[0087] Example 2

[0088] This embodiment provides a data mining system based on the industrial park platform, including a data preprocessing and feature extraction module, a model building and analysis module, and a knowledge graph integration and intelligent decision support module.

[0089] The data preprocessing and feature extraction module is used to extract features from the raw data collected from the industrial park platform and convert the raw data into useful features;

[0090] Model building and analysis module, integrating autoregressive models to predict time series data, introducing diffusion models to simulate the diffusion process of data in different dimensions, training neural networks with multiple layers and nested loop layers, optimizing model parameters through back propagation algorithms, and performing deep learning and pattern recognition of data;

[0091] The knowledge graph integration and intelligent decision support module builds and maintains the knowledge graph, including the domain knowledge and information structure of the industrial park platform, establishes the mapping relationship between the knowledge graph and the industrial park platform database, and effectively matches data and knowledge items;

[0092] Semantic relationships and query processing techniques are used to convert output data into query requests, send them to the database, and process the returned results to support intelligent decision-making and analysis.

[0093] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. The data mining method based on the industrial park platform is characterized by: The steps include: Obtain raw data from the industrial park platform and preprocess it, and use the neural network to generate new output data from the preprocessed data; Among them, the neural network is a neural network with multiple layers and nested loop layers, and the back propagation algorithm is used to train the neural network and optimize the model parameters; Based on the generated new output data, autoregressive model and diffusion model are established. Time series data are predicted through autoregressive model, and historical data points are analyzed to predict future values. Diffusion model is then used to simulate the diffusion of data in different dimensions, and the dynamic changes of data in different dimensions are analyzed and predicted. After obtaining new output data generated by the neural network, a mapping relationship between a preset knowledge graph and an industrial park platform database is established, the output data is matched with knowledge items in the preset knowledge graph, and knowledge items matching the output data are determined; Among them, the preset knowledge graph includes the domain knowledge and information structure of the industrial park platform; Before obtaining the original data from the industrial park platform, the method further includes: obtaining target data; the target data determines the data of the industrial park platform to be mined; The neural network is a neural network with multiple layers and nested loop layers, and specifically includes: The neural network includes a plurality of layers, each layer includes a set of nodes, the neural network also includes a nested loop layer, the nested loop layer includes a first loop layer and a second loop layer, the first loop layer is configured to send a message to the second loop layer; The back propagation algorithm is used to train the neural network and optimize the model parameters, specifically including: The preprocessed data is input into the neural network, and the weights and activation functions are transformed through the nodes of each layer to obtain new output data; The loss function is used to calculate the gap between the new output data of the neural network and the target value, and the loss value of the current model is obtained; According to the loss value, the gradient of each weight data of the loss function is calculated by the chain rule; wherein the gradient represents the rate of change of the loss function relative to the weight data; according to the calculated gradient, the weight data in the neural network is adjusted and updated; By iteratively training the neural network, the weight data is continuously updated to make the model output data fit the training data; Predicting time series data through autoregressive models, analyzing historical data points to predict future values, including: Determine the order of the autoregressive model and identify autocorrelation in the data by analyzing the autocorrelation function and partial autocorrelation function of time series data; Use the least squares method to estimate the parameters of the autoregressive model, including the autoregressive coefficients and the constant term; Apply the estimated parameters to the autoregressive model to fit historical data, and then perform model diagnosis; Use the fitted autoregressive model to predict future observations and predict the observations in h steps; Specifically, the value of h steps is predicted by a recursive formula, which is expressed as: in, represents the observed value predicted h steps later at time point t, c represents the model parameters, is the calculated forecast value, which is the best estimate made at time point t, is the observed value of hp predicted at time point t, p is the lag order, which is used to define the number of historical observations when the autoregressive model makes predictions; The diffusion model is used to simulate the diffusion of data in different dimensions, analyze and predict the dynamic changes of data in different dimensions. Specifically, the diffusion process of data is defined by stochastic differential equations, and the dynamic changes of data in different dimensions are captured and predicted by solving the stochastic differential equations. The specific stochastic differential equation is expressed as: in, is a multidimensional state vector, f is a drift function, is the diffusion term matrix, is the multidimensional Brownian motion vector; The establishing of a mapping relationship between a preset knowledge graph and an industrial park platform database, matching the output data with knowledge items in the preset knowledge graph, and determining knowledge items matching the output data, includes: converting the output data into a query request according to the mapping relationship; sending a query request to the industrial park platform database, and receiving a query result returned by the industrial park platform database; and using the query result as a knowledge item matching the output data; Before matching the output data with the knowledge items in the preset knowledge graph, the method further includes: obtaining the semantic relationship between the knowledge items in the preset knowledge graph; matching the output data with the knowledge items in the preset knowledge graph; After matching the output data with the knowledge items in the preset knowledge graph, it also includes: receiving a corpus associated with multiple queries from multiple users from the industrial park platform database; wherein the corpus associated with the query includes query requests and query results, and inputting the corpus into the neural network to generate an output, wherein the neural network is configured to: determine whether a word in the corpus is a stop word; when the word is not a stop word, assign a weight value to the word, and determine the importance of the word based on the assigned weight value; use a machine learning algorithm to adjust the hyperparameters of the neural network.

2. A data mining system based on an industrial park platform, using the data mining method based on an industrial park platform as claimed in claim 1, characterized in that: It includes data preprocessing and feature extraction modules, model building and analysis modules, and knowledge graph integration and intelligent decision support modules. The data preprocessing and feature extraction module is used to extract features from the original data collected from the industrial park platform; The model building and analysis module integrates an autoregressive model to predict time series data, introduces a diffusion model to simulate the diffusion process of data in different dimensions, trains a neural network with multiple layers and nested loop layers, optimizes model parameters through a back propagation algorithm, and performs deep learning and pattern recognition of data; The knowledge graph integration and intelligent decision support module constructs and maintains the knowledge graph, including the domain knowledge and information structure of the industrial park platform, establishes a mapping relationship between the knowledge graph and the industrial park platform database, and performs effective matching of data and knowledge items; It uses semantic relationships and query processing technology to convert output data into query requests, send them to the database, and process the returned results.

Images