Enterprise innovation performance prediction system for economic management

Abstract

The invention discloses an enterprise innovation performance prediction system for economic management, and the system comprises a user interface management module which is used for the login verification, data selection and result viewing of a user; the data acquisition module is used for acquiring financial data, patent data and market data through an API interface; the data processing and feature selection module is used for receiving the financial data, the patent data and the market data for data processing and feature selection to obtain data of a unified scale; the cloud model training and prediction module is used for carrying out data matching and merging based on the data of the uniform scale, carrying out model selection, model optimization and cross validation and prediction calculation, and obtaining a prediction result; the cloud server module is used for data transmission between the cloud model training and prediction module and the data acquisition module; the prediction result display module is used for displaying the prediction result in multiple dimensions; and the cloud database module is used for being connected with each module in the system and storing all feedback data in the system in real time.

Description

Technical Field

[0001] The invention belongs to the technical field of economic management and forecasting analysis, and in particular relates to an enterprise innovation performance forecasting system for economic management. Background Art

[0002] In the field of enterprise innovation performance prediction, existing technologies are mainly divided into traditional financial analysis models, econometric methods, machine learning algorithms, deep learning models, and prediction models based on text and patent data, including: (1) Traditional financial analysis models: Traditional DuPont analysis and discounted cash flow model (DCF) are widely used in enterprise value assessment. DuPont analysis decomposes the profitability and capital utilization efficiency of an enterprise through financial ratios to help managers evaluate the overall financial status of the enterprise. However, this method mainly relies on financial data and ignores the importance of intangible assets (such as patents and R&D expenditures) to innovative enterprises. The DCF model also focuses on financial data and pays less attention to the intellectual property value and innovation investment of innovative enterprises, resulting in limited application in innovation performance prediction. (2) Econometric models: Econometric methods, such as multivariate regression and structural equation model (SEM), are frequently used in enterprise performance analysis. Regression analysis is suitable for mining linear relationships in performance data, but performs poorly for nonlinear relationships commonly seen in enterprise innovation activities. Structural equation model (SEM) can handle multi-level causal relationships to a certain extent, but it is usually suitable for situations with small sample sizes and simple data structures, and has poor adaptability to high-dimensional and time series data. (3) Machine learning and data mining technology: With the popularization of machine learning technology, algorithms such as decision trees and random forests have been used to predict corporate innovation performance. These methods have the ability to automatically select features and handle nonlinear relationships, and are often used for feature importance analysis. However, random forests and decision trees are prone to overfitting when processing high-dimensional data, and it is difficult to explain the model results, which limits their interpretability in actual business applications. Support vector machines (SVMs) perform well in small samples and high-dimensional data, but their computational efficiency is low when processing large-scale data. (4) Deep learning models: In recent years, deep learning technology has gradually been applied to the performance prediction of scientific and technological innovation enterprises. For example, deep belief networks (DBNs) and convolutional neural networks (CNNs) are used to analyze time series and image data, and can capture complex nonlinear patterns in the data. However, deep learning models have "black box" properties, weak interpretability, and model training requires a lot of computing resources and data, making them difficult to adapt to small sample corporate innovation performance analysis. (5) Prediction models based on text and patent data: With the advancement of natural language processing technology, patent text analysis and technology correlation analysis are gradually being used to evaluate corporate innovation capabilities. This type of method mines the innovation potential of enterprises in the field of technology by analyzing patent content and citation networks. However, patent data analysis focuses on technical relevance and lacks the integration of corporate financial and management data, making it difficult to make a comprehensive prediction of the overall innovation performance of enterprises. In the existing technology, although there are many methods for corporate performance analysis and prediction, these methods often have certain limitations in predicting the innovation performance of high-tech enterprises.For example, traditional financial models lack the emphasis on intangible assets, machine learning models are not interpretable, and although deep learning models perform well, they require large amounts of data and lack transparency. Therefore, existing technologies still have room for improvement in accurately predicting corporate innovation performance and optimizing corporate decision support. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes an enterprise innovation performance prediction system for economic management to improve the prediction accuracy.

[0004] To achieve the above-mentioned purpose, the present invention provides an enterprise innovation performance prediction system for economic management, comprising: a user interface management module, a data acquisition module, a data processing and feature selection module, a cloud model training and prediction module, a cloud server module, a prediction result display module and a cloud database module;

[0005] The user interface management module is used for user login verification, data selection and result viewing, and is connected to the data acquisition module and the cloud database module respectively;

[0006] The data acquisition module is used to obtain financial data, patent data and market data through an API interface;

[0007] The data processing and feature selection module is used to receive the financial data, the patent data and the market data, perform data processing and feature selection, and obtain data of uniform scale;

[0008] The cloud model training and prediction module is used to perform data matching and merging, model selection, model optimization and cross-validation, and prediction calculation based on the data of the unified scale to obtain prediction results;

[0009] The cloud server module is used for data transmission between the cloud model training and prediction module and the data acquisition module;

[0010] The prediction result display module is used to display the prediction result in multiple dimensions;

[0011] The cloud database module is used to connect with various modules in the system and store all feedback data in the system in real time.

[0012] Technical effect of the invention: The invention discloses an enterprise innovation performance prediction system for economic management. Through multi-dimensional data integration and feature selection, redundant features are effectively removed, so that the model can capture the key influencing factors of enterprise innovation performance. The combined model can better handle nonlinear and complex relationships by combining machine learning and deep learning, thereby significantly improving the prediction accuracy. The combined model of the invention can adapt to diversified enterprise data, including variables of different dimensions such as finance, patents, and markets, and is suitable for the performance prediction needs of high-tech enterprises and innovation-driven enterprises. The model can also be dynamically updated so that the system maintains good adaptability when the data and market environment change. By integrating the interpretable model and integrating the bionics algorithm into the secondary stage of the integrated model to enhance the model prediction effect, the invention provides a detailed feature importance analysis and visual display, so that users can intuitively understand the prediction logic of the model. This not only improves the transparency of the prediction results, but also enhances the application value of the model in enterprise management. The invention uses natural language processing technology to combine patent text analysis with enterprise financial and market data to build a comprehensive innovation performance evaluation system. The method can more comprehensively and three-dimensionally evaluate the comprehensive performance of enterprises in technological innovation, market impact and financial status, and provide managers with all-round decision support. Through the dynamic update mechanism, the present invention can automatically adjust the model parameters to ensure the real-time and timeliness of the prediction results, and is particularly suitable for use in rapidly changing market and technological environments. Business managers can quickly adjust innovation strategies based on the latest prediction results to maintain market competitiveness. The system can help companies identify the key drivers of innovation activities and evaluate the potential for future innovation outputs, thereby optimizing resource allocation and innovation management strategies. It is not only suitable for internal corporate decision-making, but also provides investors and policymakers with innovation performance evaluation results that are of reference value. In summary, the present invention has significant advantages in improving the accuracy, applicability and transparency of innovation performance predictions, and provides companies and managers with a more scientific, real-time and comprehensive decision support tool. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0014] Figure 1 The present invention is a schematic diagram of the structure of an enterprise innovation performance prediction system for economic management according to an embodiment of the present invention. DETAILED DESCRIPTION

[0015] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0016] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0017] Problems existing in the prior art include: (1) Multidimensional data integration and feature selection: It is difficult for the prior art to effectively integrate the multidimensional data of an enterprise (such as financial data, patent data, innovation input, etc.) to comprehensively evaluate innovation performance. Traditional financial analysis models usually rely on financial data, while machine learning models often have redundant features and noise when processing multidimensional data, which affects the prediction effect. The present invention intends to achieve efficient data integration and redundant feature screening through feature selection methods (such as Lasso regression) and data cleaning to improve the accuracy and stability of the model. (2) Model adaptability and prediction accuracy: Traditional econometric models are insufficient in processing nonlinear relationships, and deep learning models are not easy to apply in the case of small samples or high-dimensional data. To this end, the present invention adopts a combined model to combine machine learning and deep learning methods, which can not only capture nonlinear features, but also effectively improve the adaptability of the model under small sample data, thereby improving prediction accuracy. (3) Model interpretability and transparency: Although the application of deep learning models (such as convolutional neural networks) in the prediction of enterprise innovation performance is effective, their "black box" nature makes the prediction results difficult to interpret, which has become an obstacle in enterprise decision-making. In order to enhance the interpretability of the model, the present invention will analyze the impact of key features based on interpretable models (such as random forests), and improve the transparency of model results and the user's trust in the prediction results through methods such as feature importance display. (4) Integrated analysis based on text and patent data: Although patent text and citation network have been studied in technological innovation evaluation, existing technologies often cannot effectively combine patent data with other aspects of the enterprise (such as financial and market data), resulting in incomplete innovation performance evaluation. The present invention intends to solve this problem by integrating the results of patent text analysis with other data levels of the enterprise through natural language processing technology and data association methods to achieve a comprehensive prediction of the overall innovation performance of the enterprise. (5) Dynamic update and real-time prediction capabilities: Most current systems are static models and cannot achieve dynamic updates of prediction models. When the external market and internal innovation input of an enterprise change, the accuracy of the prediction results will decrease over time. The present invention aims to enable the system to dynamically adapt to new data through the automatic update mechanism of the system model to ensure the real-time and timeliness of the prediction results.

[0018] like Figure 1As shown, in this embodiment, a system for predicting enterprise innovation performance for economic management is provided, including: a user interface management module, a data collection module, a data processing and feature selection module, a cloud model training and prediction module, a cloud server module, a prediction result display module and a cloud database module;

[0019] The user interface management module is used for user login verification, data selection and result viewing, and is connected to the data acquisition module and the cloud database module respectively;

[0020] The data acquisition module is used to obtain financial data, patent data and market data through an API interface;

[0021] The data processing and feature selection module is used to receive the financial data, the patent data and the market data, perform data processing and feature selection, and obtain data of uniform scale;

[0022] The cloud model training and prediction module is used to perform data matching and merging, model selection, model optimization and cross-validation, and prediction calculation based on the data of the unified scale to obtain prediction results;

[0023] The cloud server module is used for data transmission between the cloud model training and prediction module and the data acquisition module;

[0024] The prediction result display module is used to display the prediction result in multiple dimensions;

[0025] The cloud database module is used to connect with various modules in the system and store all feedback data in the system in real time.

[0026] Further, the user interface management module includes a login and authority management unit, a data input and selection unit, a prediction result display unit and a report export unit;

[0027] The login and authority management unit is used for logging in users with different roles, and each role has corresponding authority;

[0028] The data input and selection unit is used for each user to select data type and analysis dimension according to needs;

[0029] The prediction result display unit is used to display the enterprise innovation performance prediction results in real time;

[0030] The report export unit is used to export the enterprise innovation performance prediction results displayed in real time into an analysis report.

[0031] Specifically, the system supports multi-role login (such as enterprise managers, data analysts, etc.), and each role has corresponding permissions to ensure data security. Users can select data types (financial data, patent data, market data) and analysis dimensions according to their needs. The system displays the forecast results of enterprise innovation performance in real time, including trend analysis charts and feature importance analysis charts, to help users understand the key factors affecting innovation performance. Users can export analysis reports as PDF or Excel files with one click to facilitate data sharing and archiving. Interface layout: Top navigation bar: Contains quick access to login, settings, and help. Sidebar: Used to select data type and forecast time range. Main content area: Displays forecast results, feature analysis charts, and trend charts. Users can switch between different analysis charts.

[0032] Further, the data collection module includes a financial data collection unit, a patent data collection unit and a market data collection unit;

[0033] The financial data acquisition unit is used to obtain the financial data of the enterprise from the financial database through the API interface;

[0034] The patent data collection unit is used to obtain patent information from relevant departments through the API interface;

[0035] The market data collection unit is used to automatically retrieve and collect the market share and industry status of the enterprise from the market research platform through an API interface.

[0036] Furthermore, the data processing and feature selection module includes a data cleaning unit, a feature extraction unit and a normalization and standardization unit;

[0037] The data cleaning unit is used to clean the acquired financial data, patent data and market data, process missing values ​​and abnormal values ​​in the data, and obtain cleaned data;

[0038] The feature extraction unit is used to reduce the dimension of the cleaned data to obtain features after the dimension reduction, and perform feature selection on the features after the dimension reduction to obtain data after feature selection;

[0039] The normalization and standardization unit is used to perform normalization and standardization processing on the data after the feature selection to obtain data of a uniform scale.

[0040] Specifically, feature extraction (complementarity of PCA and Lasso regression):

[0041] PCA dimensionality reduction: PCA retains the main information of the data and removes collinearity by projecting the data onto a few principal components, thereby simplifying the data structure.

[0042] Lasso regression selection: Lasso regression performs sparse processing on the features extracted by PCA, reduces the weights of unimportant features to zero, and further simplifies the features. When using the LASSO regression model for feature selection, by adding a regularization penalty term to the loss function of ordinary least squares regression, the coefficients of some unimportant features are reduced to zero, thereby achieving automatic embedded feature selection. The optimization objective function of LASSO regression is shown in formula (1),

[0043]

[0044] Among them, λ is the regularization parameter, which is used to control the strength of the penalty term; β j For the j The regression coefficient of the feature; X ij For the i The sample j eigenvalues; y i For the i The true value of the samples.

[0045] Combine the complementarity of PCA and Lasso regression. PCA first performs dimensionality reduction when the data dimension is high, compressing the original feature space to fewer principal components, thereby simplifying the data structure and reducing the number of features. Then use Lasso regression to screen these principal components to further remove redundant information and find the most important features. This process can effectively reduce the computational complexity of the model while improving the generalization ability of the model. The combination of PCA and Lasso can effectively reduce data dimensions and redundant information and improve the generalization ability of the model.

[0046] Furthermore, the cloud model training and prediction module includes a cloud server data processing and matching unit, a two-stage optimization unit for the combined model, a multi-model superposition and integrated prediction unit, and a unit for selecting the optimal model for actual prediction;

[0047] The cloud server data processing and matching unit is used to build a cloud architecture, perform data matching based on the cloud architecture, and implement model updating;

[0048] The two-stage optimization unit of the combined model is used to optimize the hyperparameters of the combined model with the data after feature extraction, and perform two-stage optimization on the combined model;

[0049] The multi-model superposition and integrated prediction unit is used to perform integrated prediction by using a multi-model superposition method;

[0050] The optimal model selection unit for actual prediction is used to select the optimal model for final prediction, update data in real time, and implement model iteration.

[0051] Further, the cloud server data processing and matching unit includes a cloud architecture subunit, a data matching subunit and a model updating subunit;

[0052] The cloud architecture subunit is used to build a data warehouse in the cloud, perform data cleaning, feature selection and normalization through the cloud server, and further train and predict the cloud integrated combination model;

[0053] The data matching subunit is used for real-time data updating, data merging and matching, and feature extraction and matching based on the cleaned data;

[0054] The model updating subunit is used to update and optimize the combined model.

[0055] Specifically, cloud architecture:

[0056] Data storage: Establish a data warehouse in the cloud (such as AWS S3 or Azure Blob Storage) to store the integrated data. Data processing module: The cloud server is responsible for data cleaning, feature selection and normalization. Model training and prediction: The cloud integrates combined models (random forest, XGBoost, deep learning model, CATBoost, BP neural network) to automatically train and predict the latest data.

[0057] Data matching process:

[0058] Real-time data update: When users upload new data or regular synchronization is completed, the system triggers the data processing process. Data merging and matching: Match financial, patent and market data by enterprise ID and timestamp. Feature extraction and matching: Extract key features from cleaned data to provide effective input for the model.

[0059] Model Update:

[0060] Rolling update: The system updates the model regularly to adapt to new data and keep the prediction results up to date.

[0061] Adaptive adjustment: Use cross-validation to optimize model parameters to adapt them to dynamic environments and improve prediction accuracy.

[0062] Furthermore, the two-stage optimization unit of the combined model includes a subunit for introducing a bionic algorithm, a subunit for building a combined model, and a subunit for optimizing a bionic algorithm;

[0063] The bionic algorithm introduction subunit is used for optimizing the hyperparameters of the combined model by the system using the bionic algorithm after feature extraction;

[0064] The combined model building subunit is used to build a combined model of multiple models;

[0065] The optimization subunit of the bionic algorithm is used to run the bionic algorithm in the cloud, evaluate different parameter combinations through performance indicators, and select the best hyperparameters.

[0066] Specifically, the introduction of bionic algorithms: After feature extraction, the system uses bionic algorithms (such as genetic algorithms and particle swarm algorithms) to optimize the hyperparameters of the combined model. Bionic algorithms simulate natural selection and evolution processes, and can quickly find the optimal hyperparameter combination in a large parameter space, thereby improving the generalization ability of the model.

[0067] Construction of combined models: The system integrates multiple models, including random forest, XGBoost, deep learning models (such as CNN), CATBoost and BP neural network. Each model has its own advantages. For example, random forest and XGBoost are suitable for nonlinear feature processing, CATBoost is good at categorical variable processing, and BP neural network is suitable for mining complex nonlinear relationships.

[0068] Optimization process of bionic algorithm: The system runs the bionic algorithm in the cloud, evaluates different parameter combinations through performance indicators such as accuracy and mean square error, and selects the best hyperparameters. This process is completed automatically, and cross-validation is used to ensure the optimal performance of the final model.

[0069] Furthermore, the multi-model superposition and integrated prediction unit includes a multi-model superposition integration sub-unit and an integrated strategy sub-unit;

[0070] The multi-model superposition integration subunit is used to perform integrated prediction by using a multi-model superposition method, integrating the advantages of different models to improve prediction accuracy and robustness;

[0071] The integrated strategy subunit is used to integrate the prediction results of different models, and the bionic algorithm determines the weighting coefficient to ensure that the contribution ratio of each model reaches the optimal effect.

[0072] Furthermore, the unit for selecting the optimal model for actual prediction includes a model automatic selection and prediction subunit and a real-time update and model iteration subunit;

[0073] The model automatic selection and prediction subunit is used for the system to automatically select the best performing model for final prediction after training and optimization, so as to ensure the accuracy and adaptability of the prediction results;

[0074] The real-time update and model iteration subunit is used for the cloud server to support real-time data update. The system continuously optimizes the model parameters according to the latest data to ensure the real-time and effectiveness of the prediction results.

[0075] Further, the prediction result display module includes a prediction display unit and a system notification and feedback unit;

[0076] The forecast display unit is used to display the innovation performance forecast value, feature importance analysis and forecast report;

[0077] The system notification and feedback unit is used to automatically notify the user after generating a new prediction result, and collect feedback to further optimize the model.

[0078] Specifically, real-time prediction display:

[0079] Innovation performance forecast value: Displays the innovation performance forecast results and their changing trends through charts. Feature importance analysis: Uses SHAP values ​​to display the features that have the greatest impact on innovation performance, helping users understand the main influencing factors. Forecast report generation: Users can generate forecast reports in PDF or Excel format, including analysis charts, forecast data, and key indicators.

[0080] System notification and feedback: The system can automatically notify users after generating new prediction results and collect feedback to further optimize the model.

[0081] The advantages of the present invention over the prior art include the following aspects:

[0082] (1) Multidimensional data integration and feature selection method: This invention introduces feature selection algorithms (such as Lasso regression and weighted principal component analysis) to process multidimensional data, so as to eliminate redundant features and retain key variables that play an important role in the prediction of innovation performance. This method can effectively integrate multidimensional data such as the company's financial data, patent data, and innovation input, ensure data quality, and improve the prediction accuracy of the model. Cross-validation is used in the feature selection process to optimize the selection process and further improve the stability of the prediction results.

[0083] (2) Prediction system based on combined model: To address the nonlinear problem in enterprise innovation performance prediction, the present invention designs a combined model that combines machine learning models such as random forest and XGBoost with deep learning models (such as neural networks). In the combined model, XGBoost is used for feature engineering and data preprocessing, while the deep learning model handles complex nonlinear relationships. The combined model can effectively adapt to data of different types and sizes, thereby improving the universality and prediction accuracy of the model.

[0084] (3) Strategy to improve model interpretability: To solve the "black box" problem of deep learning models, the present invention integrates interpretable models (such as random forests) to perform feature importance analysis and visualize the impact of key features on innovation performance. Users can intuitively understand the process of forming prediction results, thereby improving the application value of the model in enterprise management. In addition, by introducing decision trees and local explanation models into the model architecture, users can better understand the prediction logic.

[0085] (4) Integrated analysis method based on text and patent data: This invention introduces natural language processing technology into patent text analysis, and converts patent information into quantifiable features through methods such as word segmentation, topic extraction, and sentiment analysis. These features are then integrated and analyzed with the company's financial data and market data to build a more comprehensive innovation performance evaluation index system. This method can comprehensively evaluate the innovation capabilities of a company from multiple perspectives such as technology, market, and finance.

[0086] (5) Real-time dynamic update mechanism: To solve the problem of prediction accuracy caused by the dynamic changes in the enterprise innovation environment and data, the present invention designs a dynamic update mechanism that allows the system to automatically adjust model parameters according to new data and external market changes. This mechanism combines rolling window technology and model automatic update algorithm, allowing the system to maintain the accuracy and timeliness of prediction results in a changing environment, meeting the enterprise's needs for real-time prediction.

[0087] The present invention discloses an enterprise innovation performance prediction system for economic management. Through multi-dimensional data integration and feature selection, redundant features are effectively removed, so that the model can capture the key influencing factors of enterprise innovation performance. The combined model can better handle nonlinear and complex relationships by combining machine learning and deep learning, thereby significantly improving the prediction accuracy. The combined model of the present invention can adapt to diversified enterprise data, including variables of different dimensions such as finance, patents, and markets, and is suitable for the performance prediction needs of high-tech enterprises and innovation-driven enterprises. The model can also be dynamically updated so that the system maintains good adaptability when the data and market environment change. By integrating an interpretable model and integrating a bionics algorithm into the secondary stage of the integrated model to enhance the model prediction effect, the present invention provides a detailed feature importance analysis and visual display, so that users can intuitively understand the prediction logic of the model. This not only improves the transparency of the prediction results, but also enhances the application value of the model in enterprise management. The present invention uses natural language processing technology to combine patent text analysis with enterprise financial and market data to build a comprehensive innovation performance evaluation system. The method can more comprehensively and three-dimensionally evaluate the comprehensive performance of enterprises in terms of technological innovation, market impact and financial status, and provide managers with all-round decision support. Through the dynamic update mechanism, the present invention can automatically adjust the model parameters to ensure the real-time and timeliness of the prediction results, and is particularly suitable for use in rapidly changing market and technological environments. Business managers can quickly adjust innovation strategies based on the latest prediction results to maintain market competitiveness. The system can help companies identify the key drivers of innovation activities and evaluate the potential for future innovation outputs, thereby optimizing resource allocation and innovation management strategies. It is not only suitable for internal corporate decision-making, but also provides investors and policymakers with innovation performance evaluation results that are of reference value. In summary, the present invention has significant advantages in improving the accuracy, applicability and transparency of innovation performance predictions, and provides companies and managers with a more scientific, real-time and comprehensive decision support tool.

[0088] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An enterprise innovation performance prediction system for economic management, characterized in that: include: User interface management module, data collection module, data processing and feature selection module, cloud model training and prediction module, cloud server module, prediction result display module and cloud database module; The user interface management module is used for user login verification, data selection and result viewing, and is connected to the data acquisition module and the cloud database module respectively; The data acquisition module is used to obtain financial data, patent data and market data through an API interface; The data processing and feature selection module is used to receive the financial data, the patent data and the market data, perform data processing and feature selection, and obtain data of a unified scale; The cloud model training and prediction module is used to perform data matching and merging, model selection, model optimization and cross-validation, and prediction calculation based on the data of the unified scale to obtain prediction results; The cloud server module is used for data transmission between the cloud model training and prediction module and the data acquisition module; The prediction result display module is used to display the prediction result in multiple dimensions; The cloud database module is used to connect with various modules in the system and store all feedback data in the system in real time.

2. The enterprise innovation performance prediction system for economic management according to claim 1, characterized in that: The user interface management module includes a login and authority management unit, a data input and selection unit, a prediction result display unit and a report export unit; The login and authority management unit is used for logging in users with different roles, and each role has corresponding authority; The data input and selection unit is used for each user to select data type and analysis dimension according to needs; The prediction result display unit is used to display the enterprise innovation performance prediction results in real time; The report export unit is used to export the real-time displayed enterprise innovation performance prediction results into an analysis report.

3. The enterprise innovation performance prediction system for economic management according to claim 1, characterized in that: The data collection module includes a financial data collection unit, a patent data collection unit and a market data collection unit; The financial data acquisition unit is used to obtain the financial data of the enterprise from the financial database through the API interface; The patent data collection unit is used to obtain patent information from relevant departments through the API interface; The market data collection unit is used to automatically retrieve and collect the market share and industry status of the enterprise from the market research platform through an API interface.

4. The enterprise innovation performance prediction system for economic management according to claim 1, characterized in that: The data processing and feature selection module includes a data cleaning unit, a feature extraction unit and a normalization and standardization unit; The data cleaning unit is used to clean the acquired financial data, patent data and market data, process missing values ​​and abnormal values ​​in the data, and obtain cleaned data; The feature extraction unit is used to reduce the dimension of the cleaned data to obtain features after the dimension reduction, and perform feature selection on the features after the dimension reduction to obtain data after feature selection; The normalization and standardization unit is used to perform normalization and standardization processing on the data after the feature selection to obtain data of a uniform scale.

5. The enterprise innovation performance prediction system for economic management according to claim 1, characterized in that: The cloud model training and prediction module includes a cloud server data processing and matching unit, a two-stage optimization unit for the combined model, a multi-model superposition and integrated prediction unit, and a unit for selecting the optimal model for actual prediction; The cloud server data processing and matching unit is used to build a cloud architecture, perform data matching based on the cloud architecture, and implement model updating; The two-stage optimization unit of the combined model is used to optimize the hyperparameters of the combined model with the data after feature extraction, and perform two-stage optimization on the combined model; The multi-model superposition and integrated prediction unit is used to perform integrated prediction by using a multi-model superposition method; The optimal model selection unit for actual prediction is used to select the optimal model for final prediction, update data in real time, and implement model iteration.

6. The enterprise innovation performance prediction system for economic management according to claim 5, characterized in that: The cloud server data processing and matching unit includes a cloud architecture subunit, a data matching subunit and a model updating subunit; The cloud architecture subunit is used to build a data warehouse in the cloud, perform data cleaning, feature selection and normalization through the cloud server, and further train and predict the cloud integrated combination model; The data matching subunit is used for real-time data updating, data merging and matching, and feature extraction and matching based on the cleaned data; The model updating subunit is used to update and optimize the combined model.

7. The enterprise innovation performance prediction system for economic management according to claim 5, characterized in that: The two-stage optimization unit of the combined model includes a bionic algorithm introduction subunit, a combined model construction subunit and a bionic algorithm optimization subunit; The bionic algorithm introduction subunit is used for optimizing the hyperparameters of the combined model by the system using the bionic algorithm after feature extraction; The combined model building subunit is used to build a combined model of multiple models; The optimization subunit of the bionic algorithm is used to run the bionic algorithm in the cloud, evaluate different parameter combinations through performance indicators, and select the best hyperparameters.

8. The enterprise innovation performance prediction system for economic management according to claim 5, characterized in that: The multi-model superposition and integrated prediction unit includes a multi-model superposition integration sub-unit and an integrated strategy sub-unit; The multi-model superposition integration subunit is used to perform integrated prediction by using a multi-model superposition method, integrating the advantages of different models to improve prediction accuracy and robustness; The integrated strategy subunit is used to integrate the prediction results of different models, and the bionic algorithm determines the weighting coefficient to ensure that the contribution ratio of each model reaches the optimal effect.

9. The enterprise innovation performance prediction system for economic management according to claim 5, characterized in that: The unit for selecting the optimal model for actual prediction includes a model automatic selection and prediction subunit and a real-time update and model iteration subunit; The model automatic selection and prediction subunit is used for the system to automatically select the best performing model for final prediction after training and optimization, so as to ensure the accuracy and adaptability of the prediction results; The real-time update and model iteration subunit is used for the cloud server to support real-time data update. The system continuously optimizes the model parameters according to the latest data to ensure the real-time and effectiveness of the prediction results.

10. The enterprise innovation performance prediction system for economic management according to claim 1, characterized in that: The prediction result display module includes a prediction display unit and a system notification and feedback unit; The forecast display unit is used to display the innovation performance forecast value, feature importance analysis and forecast report; The system notification and feedback unit is used to automatically notify the user after generating a new prediction result, and collect feedback to further optimize the model.

Images