Intelligent assessment and transformation management system for scientific and technological achievements

Abstract

The invention discloses an intelligent assessment and transformation management system for scientific and technological achievements, and belongs to the technical field, and in the intelligent assessment and transformation management system, the combination of DNN and RNN provides powerful feature learning and representation capabilities for the system. The DNN can extract high-level features from complex data of scientific and technological achievements, and the RNN can capture a long-term dependency relationship in time sequence data. And meanwhile, the XGBoost model algorithm can utilize the features extracted by the DNN and the RNN and the original features, and through constructing a plurality of decision trees and carrying out gradient lifting, the accuracy and reliability of prediction are improved, the nonlinear relationship between the features is effectively processed, the model deviation is reduced, and the evaluation result of the scientific and technological achievements is more accurate. The combination of the three algorithms enables the advantages of each model to be complementary, and the deep feature extraction of the DNN, the time sequence processing capability of the RNN and the integrated learning mechanism of the XGBoost act together, so that the over-fitting risk can be effectively reduced. The diversity enables the model to have better generalization ability when facing unknown data.

Description

Technical Field

[0001] The present invention belongs to the technical field of evaluation of scientific and technological achievements transformation, and specifically relates to an intelligent evaluation and transformation management system for scientific and technological achievements. Background Art

[0002] The evaluation and transformation of scientific and technological achievements are important links in promoting scientific and technological innovation and industrial development. It involves the value judgment of scientific and technological achievements, market prospect analysis, and the whole process of transforming them into actual productivity. The following is a brief introduction to the evaluation and transformation of scientific and technological achievements: The evaluation of scientific and technological achievements refers to the comprehensive evaluation of scientific and technological achievements completed by scientific research institutions or individuals in terms of scientificity, innovation, practicability, maturity, and market prospects. The purpose of the evaluation is to determine the value of the achievements and provide a basis for the transformation of the achievements.

[0003] Although some evaluation models based on artificial intelligence have been applied to the field of scientific and technological achievements evaluation, most of these models use a single algorithm, such as deep neural network (DNN) or gradient-boosted decision tree (XGBoost), and it is difficult to comprehensively capture the multi-dimensional features and time series information of scientific and technological achievements. In addition, the existing systems have not formed a complete, efficient, and intelligent management system in terms of data collection and processing, data storage and backup, and the design of the achievement transformation module, and cannot meet the needs of modern scientific and technological achievements evaluation and transformation. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent evaluation and transformation management system for scientific and technological achievements in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: An intelligent evaluation and transformation management system for scientific and technological achievements, the system includes: a basic database module, a data collection and processing module, a data storage and backup module, a data layer module, a service layer module, an application layer module, and a user interface layer module;

[0006] An intelligent evaluation module, an achievement transformation module, and an expert review module are provided inside the service layer module;

[0007] The data collection and processing module is responsible for collecting relevant information of scientific and technological achievements from various data sources, performing preliminary data cleaning and formatting, and then transmitting the processed data to the data storage and backup module. The data storage and backup module is responsible for storing the collected data in the database

[0008] The service layer module is built on top of the data layer. It obtains the required data by calling the data access interfaces provided by the data layer and implements intelligent evaluation algorithms to evaluate scientific and technological achievements. The application layer module is responsible for implementing the business logic of the system. It presents the data and functions provided by the service layer to users in specific application forms, such as generating evaluation reports and providing suggestions for achievement transformation.

[0009] The user interface layer module is the direct window for users to interact with the system. It presents the services provided by the application layer to users in the form of a graphical interface and collects users' inputs and requests.

[0010] In a preferred embodiment, the basic database module includes three main sub-libraries: the scientific and technological achievements library, the expert library, and the policy and regulation library. Each sub-library undertakes specific data storage tasks.

[0011] The scientific and technological achievements library is the core of the basic database module. It stores in detail the detailed information of various scientific and technological achievements, including key information such as the name, type, field of belonging, R & D unit, patent situation, technical maturity, and market potential of the achievements. These data are important bases for evaluating and transforming scientific and technological achievements, providing rich achievement resources for the system and facilitating subsequent analysis and decision-making.

[0012] The expert library focuses on storing expert information related to the evaluation and transformation of scientific and technological achievements, including the basic personal information, professional fields, research achievements, review experiences, and academic status of experts. This library provides an expert resource pool for the system, which can provide professional opinions and suggestions during the evaluation process. At the same time, it can also match corresponding experts for review and guidance according to the fields and characteristics of scientific and technological achievements.

[0013] The policy and regulation library is the legal support part of the basic database module. It collects and stores relevant national and local policies, regulations, industry standards and other documents. These policy and regulation information is of great significance for the compliance evaluation, market access, and intellectual property protection of scientific and technological achievements, providing policy bases and guidance for the transformation of scientific and technological achievements.

[0014] In a preferred embodiment, the data collection and processing module is the front-end module of the intelligent evaluation and transformation management system for scientific and technological achievements, responsible for collecting relevant information on scientific and technological achievements from multiple data sources (such as scientific research databases, patent databases, academic journals, technical reports, etc.). The specific content and operation method of this module include: First, regularly capture data through automated scripts or API interfaces, etc.; then, preprocess the collected data, including data cleaning (removing invalid and incomplete data), data formatting (unifying data formats and units), data conversion (such as encoding categorical data into numerical data), etc.; finally, transfer the cleaned and formatted data to the data storage and backup module for further processing and storage. This module is a key step to ensure the accuracy of subsequent analysis, and it provides a high-quality data foundation for the system through an efficient data processing process.

[0015] In a preferred embodiment, the data storage and backup module is the core of the system's data management, responsible for securely and effectively storing the data transmitted by the data collection and processing module in the database. The specific content and operation method of this module include: Select a suitable database system (such as the relational database MySQL, the non-relational database MongoDB, etc.) to store data, and design a reasonable database structure to optimize data access efficiency; implement data writing and reading operations to ensure data integrity and consistency; regularly perform data backup operations to prevent data loss or damage. The backup can be at different locations on the same server or a remote cloud storage service; at the same time, this module is also responsible for database maintenance and performance monitoring to ensure the stable operation of the data storage system.

[0016] In a preferred embodiment, the data layer module directly interacts with the database and provides data access and management services for the upper-layer modules. To ensure the persistent storage and fast retrieval of data, the operation method of the data layer module includes the following steps:

[0017] Initialization: When the system starts up, the data layer module initializes the database connection pool and establishes a connection with the database.

[0018] Request processing: When the service layer or application layer module needs to access data, they send requests through the interfaces provided by the data layer module.

[0019] Operation execution: After receiving the request, the data layer module executes the corresponding database operations, such as querying, updating, inserting, or deleting data.

[0020] Result return: After the operation is completed, the data layer module returns the result to the requester or processes exception situations as needed.

[0021] Resource release: After the operation is completed, the data layer module is responsible for releasing the database connection resources for subsequent requests to use.

[0022] In a preferred embodiment, the intelligent evaluation module employs three models: Deep Neural Network (DNN), Recurrent Neural Network (RNN), and eXtreme Gradient Boosting (XGBoost). First, through its multi-layer perception structure, the Deep Neural Network learns high-level abstract features from the scientific and technological achievement data. The Recurrent Neural Network processes time series data or information with sequence dependence, such as the development process and stage achievements of scientific and technological achievements. The RNN can capture long-term dependencies in the time series and provide temporal features for evaluation. The prediction model of eXtreme Gradient Boosting (XGBoost) fuses the features extracted by the DNN and RNN with the original features, and through constructing multiple decision trees and performing gradient boosting, it realizes the comprehensive evaluation and prediction of scientific and technological achievements, specifically including:

[0023] S1. Data preprocessing: Data cleaning, deleting null values, outliers, and duplicate records.

[0024] S2. Feature selection: Use the DNN model to evaluate the feature importance by observing the contribution of each feature to the model loss. According to the feature importance score, select the top N features

[0025] S3. Model training

[0026] S3-1. Data partitioning, dividing the dataset into a training set, a validation set, and a test set.

[0027] S3-2. DNN model construction, designing the DNN architecture including: an input layer, multiple hidden layers, and an output layer.

[0028] S3-3. RNN model construction, first designing the RNN architecture: using LSTM or GRU units to process sequence data. The formula for the LSTM unit is:

[0029] Forget gate formula: f t =σ(W f ·[h t-1 ,x t +b f );

[0030] Where: ft represents the output of the forget gate at time step t.

[0031] Wf represents the weight matrix of the forget gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the forget gate.

[0032] [ht-1,xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, serving as the input to the forget gate.

[0033] bf represents the bias term of the forget gate, which is used to adjust the output of the forget gate.

[0034] Input gate formula: i t = σ(W i · [h t-1 , x t + b i );

[0035] where: it represents the output of the input gate at time step t.

[0036] Wi represents the weight matrix of the input gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the input gate.

[0037] [ht-1, xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, which serves as the input to the input gate.

[0038] bi represents the bias term of the input gate, which is used to adjust the output of the input gate.

[0039] New candidate value formula:

[0040] where:

[0041] C~t: The candidate value, which is the new information that may be added to the cell state at the current time step t.

[0042] WC represents the weight matrix of the candidate value, which determines the influence of the previous hidden state ht-1 and the current input xt on the candidate value.

[0043] [ht-1, xt]: The concatenation of the previous hidden state ht-1 and the current input xt, which serves as the input to the candidate value.

[0044] bC represents the bias term of the candidate value, which is used to adjust the output of the candidate value.

[0045] Cell state update formula:

[0046] where: Ct represents the cell state at time step t, which is the memory of the LSTM cell.

[0047] ft represents the output of the forget gate, which determines how to forget or retain the previous cell state Ct-1.

[0048] Ct-1 represents the cell state of the previous time step.

[0049] it represents the output of the input gate, which determines how to add the candidate value C~t to the cell state

[0050] Output gate formula: o t = σ(Wo · [h t-1 , x t + b o );

[0051] Where: ot represents the output of the output gate at time step t.

[0052] Wo represents the weight matrix of the output gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the output gate.

[0053] [ht-1, xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, serving as the input to the output gate.

[0054] bo represents the bias term of the output gate, used to adjust the output of the output gate

[0055] Hidden state formula:

[0056] Where: ht: The hidden state at time step t, which is the output of the LSTM cell.

[0057] ot represents the output of the output gate, which determines which parts of the cell state Ct will be output.

[0058] tanh(Ct) represents the processing of the cell state Ct through the hyperbolic tangent function, compressing the values into the interval [-1, 1]

[0059] S3-4. Initialize the XGBoost model parameters, including the number of trees, the depth of the trees, the learning rate, etc.

[0060] S3-5. Model training: Use the training set to train the DNN, RNN, and XGBoost models respectively. DNN training: Update the weights through backpropagation and gradient descent. RNN training: Also use backpropagation, considering the gradients of the time series. XGBoost training: Use the gradient boosting framework to build decision trees and optimize the objective function;

[0061] S4. Model fusion, use the DNN and RNN models to perform feature transformation on the training set data, obtain the output of the hidden layer as the new feature representation, combine the output features of the DNN and RNN models with the original features to form a new feature set, and then use the fused feature set to train the XGBoost model; where the feature fusion formula is:

[0062] X_{fusion} = [X_{DNN}, X_{RNN}, X_{original}]

[0063] S5. Model Application: The samples to be evaluated are subjected to feature transformation through DNN and RNN models. The transformed features are input into the XGBoost model for final prediction.

[0064] In a preferred embodiment, the achievement transformation module is the core execution module of the intelligent evaluation and transformation management system for scientific and technological achievements. It is responsible for promoting the evaluated scientific and technological achievements to the market and realizing their commercial value. The specific content and operation method of this module include: First, by analyzing the comprehensive evaluation results of scientific and technological achievements, determine their transformation potential and suitable transformation paths; Second, the module will automatically match potential cooperative enterprises or investment institutions and recommend the optimal partners through algorithms; Then, the module provides a complete set of transformation process management tools, including project planning, fund raising, team building, market promotion, etc.; During this process, the module will monitor the progress and effect of the transformation in real time, and feedback the problems and risks in the transformation process through data analysis; Finally, the achievement transformation module will also provide follow-up services after the transformation is successful, including market feedback collection, subsequent R & D support, intellectual property management, etc., to ensure that scientific and technological achievements can continuously generate benefits. This module is the bridge connecting scientific research and the market. It promotes scientific and technological achievements from theory to practice and realizes the value of scientific and technological innovation through efficient management and optimized processes.

[0065] In a preferred embodiment, for the expert review module, first, according to the field and characteristics of scientific and technological achievements, the system will screen out experts with corresponding professional backgrounds and review experience from the expert database; Second, the module will automatically send review invitations to the selected experts and provide them with an online review platform to facilitate experts to conduct review work anytime and anywhere; Next, the experts comprehensively evaluate the originality, practicality, market prospects, etc. of scientific and technological achievements through the platform and give specific scores and suggestions; The module will collect the review results of all experts and obtain the final review report through algorithm analysis; Finally, the module will feedback the review report to the submitters of scientific and technological achievements and provide a reference for the decision-making of the management level. The expert review module ensures the fairness and authority of the evaluation of scientific and technological achievements through professional and standardized review processes and provides an important reference basis for the transformation of scientific and technological achievements.

[0066] In a preferred embodiment, the application layer module is the business logic implementation layer of the intelligent evaluation and transformation management system for scientific and technological achievements. It presents the data and functions provided by the service layer to users in a specific application form. The specific content and operation methods of this module include: defining various business processes in the system, such as the achievement evaluation process, the transformation suggestion generation process, etc.; calling the API interfaces provided by the service layer to execute specific business operations, such as generating evaluation reports, providing transformation path suggestions, etc.; processing users' business requests and converting the processing results of the service layer into formats understandable by users, such as charts, documents, etc.; at the same time, the application layer is also responsible for the security, transaction management, and session management of the system to ensure the reliability of user operations and the security of data.

[0067] In a preferred embodiment, the user interface layer module is the direct window for users to interact with the intelligent evaluation and transformation management system for scientific and technological achievements. It is responsible for presenting the services provided by the application layer to users in an intuitive graphical interface form. The specific content and operation methods of this module include: designing a user-friendly interface to enable users to easily operate and obtain information; implementing the input and output of user operations, such as form submission, button clicking, information display, etc.; collecting users' inputs and requests and passing them to the application layer for processing; receiving the processing results of the application layer and presenting them to users in the form of charts, lists, reports, etc.; in addition, the user interface layer is also responsible for implementing functions such as user authentication and permission control to ensure the security of the system and the privacy protection of users.

[0068] In summary, due to the adoption of the above technical solutions, the beneficial effects of the present invention are as follows:

[0069] 1. In the present invention, the combination of DNN and RNN provides the system with powerful feature learning and representation capabilities. DNN can extract high-level features from the complex data of scientific and technological achievements, while RNN can capture long-term dependencies in time series data. This combination enables the system to more comprehensively understand the inherent characteristics of scientific and technological achievements, including their historical evolution and future potential. At the same time, the XGBoost model algorithm can utilize the features extracted by DNN and RNN as well as the original features to improve the accuracy and reliability of prediction by constructing multiple decision trees and performing gradient boosting. Thus, it can effectively handle the non-linear relationships between features, reduce model bias, and make the evaluation results of scientific and technological achievements more accurate. The combination of the three algorithms complements the advantages of each model. The deep feature extraction of DNN, the time series processing ability of RNN, and the ensemble learning mechanism of XGBoost work together to effectively reduce the risk of overfitting. This diversity enables the model to have better generalization ability when facing unknown data.

[0070] 2. In the present invention, the data acquisition and processing module ensures that the system efficiently and accurately acquires and preprocesses data from multiple data sources, providing a high-quality data basis for subsequent analysis. Secondly, the data storage and backup module guarantees the security and accessibility of data, providing support for the stable operation of the system and the long-term preservation of data. Then, the intelligent evaluation module improves the accuracy and efficiency of scientific and technological achievement evaluation through advanced algorithm models, providing a scientific reference basis for decision-makers. In addition, the achievement transformation module promotes the docking of scientific and technological achievements with market demands and accelerates the commercialization process of scientific and technological achievements. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 is the overall system block diagram of the present invention;

[0072] Figure 2 is the system block diagram of the service layer module in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0073] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0074] Refer to Figure 1-2 ,

[0075] Embodiment:

[0076] An intelligent evaluation and transformation management system for scientific and technological achievements, the system includes: a basic database module, a data acquisition and processing module, a data storage and backup module, a data layer module, a service layer module, an application layer module and a user interface layer module;

[0077] The service layer module internally is provided with an intelligent evaluation module, an achievement transformation module and an expert review module;

[0078] The data acquisition and processing module is responsible for collecting relevant information on scientific and technological achievements from various data sources, and performing preliminary data cleaning and formatting, and then transferring the processed data to the data storage and backup module. The data storage and backup module is responsible for storing the collected data in the database

[0079] The service layer module is built on the data layer, obtains the required data by calling the data access interface provided by the data layer, and implements an intelligent evaluation algorithm to evaluate scientific and technological achievements. The application layer module is responsible for implementing the business logic of the system, and it presents the data and functions provided by the service layer to the user in specific application forms, such as evaluation report generation, achievement transformation suggestions, etc.

[0080] The user interface layer module is the direct window for users to interact with the system. It presents the services provided by the application layer to users in the form of a graphical interface and collects users' inputs and requests.

[0081] The basic database module contains three main sub - databases: the scientific and technological achievements database, the expert database, and the policies and regulations database. Each sub - database undertakes specific data storage tasks.

[0082] The scientific and technological achievements database is the core of the basic database module. It stores in detail the detailed information of various scientific and technological achievements, including key information such as the name, type, field of belonging, R & D unit, patent situation, technological maturity, and market potential of the achievements. These data are important bases for the evaluation and transformation of scientific and technological achievements, providing rich achievement resources for the system and facilitating subsequent analysis and decision - making.

[0083] The expert database focuses on storing expert information related to the evaluation and transformation of scientific and technological achievements, including the basic personal information, professional fields, research achievements, review experiences, and academic status of experts. This database provides an expert resource pool for the system, which can provide professional opinions and suggestions during the evaluation process. At the same time, it can also match corresponding experts for review and guidance according to the fields and characteristics of scientific and technological achievements.

[0084] The policies and regulations database is the legal support part of the basic database module. It collects and stores relevant national and local policies, regulations, industry standards and other documents. These policy and regulation information is of great significance for the compliance evaluation, market access, and intellectual property protection of scientific and technological achievements, providing policy bases and guidance for the transformation of scientific and technological achievements.

[0085] The data collection and processing module is the front - end module of the intelligent evaluation and transformation management system for scientific and technological achievements. It is responsible for collecting relevant information of scientific and technological achievements from multiple data sources (such as scientific research databases, patent databases, academic journals, technical reports, etc.). The specific content and operation methods of this module include: First, regularly capture data through automated scripts or API interfaces, etc.; then, pre - process the collected data, including data cleaning (removing invalid and incomplete data), data formatting (unifying data formats and units), data conversion (such as encoding categorical data into numerical data), etc.; finally, transfer the cleaned and formatted data to the data storage and backup module for further processing and storage. This module is a key step to ensure the accuracy of subsequent analysis. It provides a high - quality data foundation for the system through an efficient data processing process.

[0086] The data storage and backup module is the core of the system's data management, responsible for securely and effectively storing the data passed by the data collection and processing module in the database. The specific content and operation methods of this module include: selecting a suitable database system (such as the relational database MySQL, the non-relational database MongoDB, etc.) to store data, and designing a reasonable database structure to optimize data access efficiency; implementing data writing and reading operations to ensure data integrity and consistency; regularly performing data backup operations to prevent data loss or damage, and the backup can be at different locations on the same server or a remote cloud storage service; at the same time, this module is also responsible for database maintenance and performance monitoring to ensure the stable operation of the data storage system.

[0087] The data layer module directly interacts with the database and provides data access and management services for the upper-layer modules. To ensure the persistent storage and fast retrieval of data, the operation methods of the data layer module include the following steps:

[0088] Initialization: When the system starts up, the data layer module initializes the database connection pool and establishes a connection with the database.

[0089] Request processing: When the service layer or application layer module needs to access data, they send requests through the interfaces provided by the data layer module.

[0090] Execute operations: After receiving the request, the data layer module executes the corresponding database operations, such as querying, updating, inserting, or deleting data.

[0091] Return results: After the operation is completed, the data layer module returns the results to the requester or handles exception situations as needed.

[0092] Resource release: After the operation is completed, the data layer module is responsible for releasing the database connection resources for subsequent requests to use.

[0093] The intelligent evaluation module adopts three models: the deep neural network (DNN), the recurrent neural network (RNN), and the gradient-boosted decision tree (XGBoost). First, the deep neural network learns high-level abstract features from the scientific and technological achievement data through its multi-layer perceptron structure. The recurrent neural network processes time series data or information with sequence dependencies, such as the development process and stage achievements of scientific and technological achievements. The RNN can capture long-term dependencies in the time series and provide temporal features for evaluation. The gradient-boosted decision tree XGBoost prediction model fuses the features extracted by the DNN and RNN with the original features, and realizes the comprehensive evaluation and prediction of scientific and technological achievements by constructing multiple decision trees and performing gradient boosting. Specifically, it includes:

[0094] S1. Data preprocessing: Data cleaning, deleting null values, outliers, and duplicate records.

[0095] S2. Feature Selection: Use the DNN model to evaluate feature importance by observing the contribution of each feature to the model loss. Select the top N features according to the feature importance scores.

[0096] S3. Model Training

[0097] S3-1. Data Partitioning: Divide the dataset into a training set, a validation set, and a test set.

[0098] S3-2. DNN Model Construction: Design the DNN architecture including an input layer, multiple hidden layers, and an output layer.

[0099] S3-3. RNN Model Construction: First, design the RNN architecture: Use LSTM or GRU units to process sequential data. The LSTM unit formula is:

[0100] Forget gate formula: f t = σ(W f · [h t-1 , x t + b f );

[0101] where: ft represents the output of the forget gate at time step t.

[0102] Wf represents the weight matrix of the forget gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the forget gate.

[0103] [ht-1, xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, which serves as the input to the forget gate.

[0104] bf represents the bias term of the forget gate, which is used to adjust the output of the forget gate.

[0105] Input gate formula: i t = σ(W i · [h t-1 , x t + b i );

[0106] where: it represents the output of the input gate at time step t.

[0107] Wi represents the weight matrix of the input gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the input gate.

[0108] [ht-1, xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, which serves as the input to the input gate.

[0109] bi represents the bias term of the input gate, which is used to adjust the output of the input gate.

[0110] New candidate value formula:

[0111] Where:

[0112] C~t: Candidate value, which is the new information that may be added to the cell state at the current time step t.

[0113] WC: Weight matrix of the candidate value, which determines the influence of the previous hidden state ht-1 and the current input xt on the candidate value.

[0114] [ht-1, xt]: Concatenation of the previous hidden state ht-1 and the current input xt, serving as the input to the candidate value.

[0115] bC: Bias term of the candidate value, used to adjust the output of the candidate value.

[0116] Cell state update formula:

[0117] Where Ct represents the cell state at time step t and is the memory of the LSTM cell.

[0118] ft represents the output of the forget gate, which determines how to forget or retain the previous cell state Ct-1.

[0119] Ct-1 represents the cell state of the previous time step.

[0120] it represents the output of the input gate, which determines how to add the candidate value C~t to the cell state

[0121] Output gate formula: o t = σ(W o ·[h t-1 , x t +b o );

[0122] Where ot represents the output of the output gate at time step t.

[0123] Wo represents the weight matrix of the output gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the output gate.

[0124] [ht-1, xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, serving as the input to the output gate.

[0125] bo represents the bias term of the output gate, used to adjust the output of the output gate

[0126] Hidden state formula:

[0127] where: ht: the hidden state at time step t, which is the output of the LSTM cell.

[0128] ot represents the output of the output gate, which determines which parts of the cell state Ct will be output.

[0129] tanh(Ct) represents the processing of the cell state Ct through the hyperbolic tangent function, which compresses the values into the interval [-1, 1].

[0130] S3-4. Initialize the XGBoost model parameters, including the number of trees, the depth of the trees, the learning rate, etc.

[0131] S3-5. Model training: Use the training set to train the DNN, RNN, and XGBoost models respectively. DNN training: Update the weights through backpropagation and gradient descent. RNN training: Also use backpropagation, considering the gradients of the time series. XGBoost training: Use the gradient boosting framework to build decision trees and optimize the objective function.

[0132] S4. Model fusion: Use the DNN and RNN models to perform feature transformation on the training set data, obtain the output of the hidden layer as a new feature representation, combine the output features of the DNN and RNN models with the original features to form a new feature set, and then use the fused feature set to train the XGBoost model; where the feature fusion formula is:

[0133] X_{fusion} = [X_{DNN}, X_{RNN}, X_{original}]

[0134] S5. Model application: Perform feature transformation on the samples to be evaluated through the DNN and RNN models. Input the transformed features into the XGBoost model for final prediction.

[0135] The achievement transformation module is the core execution module of the intelligent evaluation and transformation management system for scientific and technological achievements. It is responsible for promoting the evaluated scientific and technological achievements to the market and realizing their commercial value. The specific content and operation methods of this module include: First, by analyzing the comprehensive evaluation results of scientific and technological achievements, determine their transformation potential and appropriate transformation paths; Second, the module will automatically match potential cooperative enterprises or investment institutions and recommend the optimal partners through algorithms; Then, the module provides a complete set of transformation process management tools, including project planning, fund raising, team building, market promotion, etc.; During this process, the module will monitor the progress and effect of the transformation in real time, and feedback the problems and risks in the transformation process through data analysis; Finally, the achievement transformation module will also provide follow-up services after the transformation is successful, including market feedback collection, subsequent R & D support, intellectual property management, etc., to ensure that scientific and technological achievements can continuously generate benefits. This module is the bridge connecting scientific research and the market. It promotes the transformation of scientific and technological achievements from theory to practice and realizes the value of scientific and technological innovation through efficient management and optimized processes.

[0136] For the expert review module, first, according to the fields and characteristics of scientific and technological achievements, the system will screen out experts with corresponding professional backgrounds and review experiences from the expert database; Second, the module will automatically send review invitations to the selected experts and provide them with an online review platform to facilitate experts to conduct review work anytime and anywhere; Next, the experts will comprehensively evaluate the originality, practicality, market prospects, etc. of the scientific and technological achievements through the platform and give specific scores and suggestions; The module will collect the review results of all experts and obtain the final review report through algorithm analysis; Finally, the module will feedback the review report to the submitters of scientific and technological achievements and provide a reference for the decision-making of the management level at the same time. The expert review module ensures the fairness and authority of the evaluation of scientific and technological achievements through professional and standardized review processes and provides an important reference basis for the transformation of scientific and technological achievements.

[0137] The application layer module is the business logic implementation layer of the intelligent evaluation and transformation management system for scientific and technological achievements. It presents the data and functions provided by the service layer in specific application forms to users. The specific content and operation methods of this module include: defining various business processes in the system, such as the achievement evaluation process, the transformation suggestion generation process, etc.; calling the API interfaces provided by the service layer to execute specific business operations, such as generating evaluation reports, providing transformation path suggestions, etc.; processing users' business requests and converting the processing results of the service layer into formats understandable by users, such as charts, documents, etc.; At the same time, the application layer is also responsible for the security, transaction management and session management of the system to ensure the reliability of user operations and the security of data.

[0138] The user interface layer module is the direct window for users to interact with the intelligent evaluation and transformation management system of scientific and technological achievements. It is responsible for presenting the services provided by the application layer to users in the form of an intuitive graphical interface. The specific content and operation methods of this module include: designing a user-friendly interface to enable users to easily operate and obtain information; implementing the input and output of user operations, such as form submission, button clicks, information display, etc.; collecting user inputs and requests and passing them to the application layer for processing; receiving the processing results from the application layer and presenting them to users in the form of charts, lists, reports, etc.; in addition, the user interface layer is also responsible for implementing functions such as user authentication and permission control to ensure the security of the system and the privacy protection of users.

[0139] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0140] The above description enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. Intelligent evaluation and transformation management system for scientific and technological achievements, characterized by: The system includes: a basic database module, a data collection and processing module, a data storage and backup module, a data layer module, a service layer module, an application layer module and a user interface layer module; The service layer module is internally provided with an intelligent evaluation module, an achievement transformation module and an expert review module; The data collection and processing module is responsible for collecting relevant information of scientific and technological achievements from various data sources, and performing preliminary data cleaning and formatting, and then passing the processed data to the data storage and backup module; the data storage and backup module is responsible for storing the collected data in the database The service layer module is built on the data layer, obtains the required data by calling the data access interface provided by the data layer, and implements the intelligent evaluation algorithm to evaluate the scientific and technological achievements; the application layer module is responsible for implementing the business logic of the system, and it presents the data and functions provided by the service layer to users in a specific application form; The user interface layer module is a direct window for users to interact with the system. It displays the services provided by the application layer to users in the form of a graphical interface and collects user input and requests.

2. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The basic database module includes three main sub-databases: scientific and technological achievements database, expert database and policy and regulations database, each of which undertakes specific data storage tasks; The scientific and technological achievements database is the core of the basic database module. It stores detailed information on various scientific and technological achievements, including the name, type, field, R&D unit, patent status, technology maturity, and market potential of the achievements. These data are important bases for the evaluation and transformation of scientific and technological achievements, providing the system with rich achievement resources and facilitating subsequent analysis and decision-making. The expert database is dedicated to storing expert information related to the evaluation and transformation of scientific and technological achievements, including the expert's basic personal information, professional fields, research results, review experience, and academic status; this database provides the system with an expert resource pool, providing professional opinions and suggestions during the evaluation process, and can also match corresponding experts for review and guidance according to the fields and characteristics of scientific and technological achievements; The policy and regulations database is the legal support part of the basic database module, which collects and stores relevant national and local policies, regulations, and industry standard documents; These policy and regulatory information are of great significance for the compliance assessment, market access, and intellectual property protection of scientific and technological achievements, and provide policy basis and guidance for the transformation of scientific and technological achievements.

3. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The data acquisition and processing module is the front-end module of the intelligent evaluation and transformation management system of scientific and technological achievements, which is responsible for collecting relevant information of scientific and technological achievements from multiple data sources; the specific content and operation method of this module include: first, regularly capturing data through automated scripts or API interfaces; then, pre-processing the collected data, including data cleaning, data formatting, and data conversion steps; finally, passing the cleaned and formatted data to the data storage and backup module for further processing and storage; this module is a key step to ensure the accuracy of subsequent analysis, and it provides the system with a high-quality data foundation through an efficient data processing process.

4. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The data storage and backup module is the data management core of the system, responsible for storing the data transmitted by the data acquisition and processing module in the database safely and effectively; the specific content and operation method of this module include: selecting a suitable database system to store data, and designing a reasonable database structure to optimize data access efficiency; implementing data writing and reading operations to ensure data integrity and consistency; regularly performing data backup operations to prevent data loss or damage. The backup is in different locations on the same server, and it is also a remote cloud storage service; at the same time, this module is also responsible for database maintenance and performance monitoring to ensure the stable operation of the data storage system.

5. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The data layer module directly interacts with the database to provide data access and management services for the upper layer module, ensuring persistent storage and fast retrieval of data. The data layer module operation method includes the following steps: Initialization: When the system starts, the data layer module initializes the database connection pool and establishes a connection with the database; Request processing: When the service layer or application layer modules need to access data, they send requests through the interface provided by the data layer module; Execute operations: After receiving the request, the data layer module performs the corresponding database operations, including querying, updating, inserting or deleting data; Return result: After the operation is completed, the data layer module returns the result to the requester, or handles abnormal situations as needed; Resource release: After the operation is completed, the data layer module is responsible for releasing the database connection resources for subsequent requests.

6. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The intelligent evaluation module adopts three models: deep neural network, recurrent neural network and decision tree based on gradient boosting. First, the deep neural network learns high-level abstract features from scientific and technological achievement data through its multi-layer perception structure, and the recurrent neural network processes time series data or information with sequence dependencies, including: S1. Data preprocessing: data cleaning, deletion of null values, outliers and duplicate records; S2. Feature selection: Use the DNN model to evaluate feature importance by observing the contribution of each feature to the model loss; select the top N features based on the feature importance score S3. Model training S3-1. Data division: divide the data set into training set, validation set and test set; S3-2.DNN model construction, design DNN architecture including: input layer, multiple hidden layers, and output layer; S3-3.RNN model construction, first design the RNN architecture: use LSTM or GRU units to process sequence data, the LSTM unit formula is: Forget gate formula: f t =σ(W f ·[h t-1 , x t ]+b f ); Where: ft represents the output of the forget gate at time step t; Wf represents the weight matrix of the forget gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the forget gate; [ht-1,xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, which serves as the input of the forget gate; bf represents the bias term of the forget gate, which is used to adjust the output of the forget gate; Input gate formula: i t =σ(W i ·[h t-1 ,x t ]+b i ); Where: it represents the output of the input gate at time step t; Wi represents the weight matrix of the input gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the input gate; [ht-1,xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, which serves as the input of the input gate; bi represents the bias term of the input gate, which is used to adjust the output of the input gate; New candidate value formula: in: C~t: Candidate value, which is the new information that may be added to the cell state at the current time step t; WC: The weight matrix of the candidate value, which determines the influence of the previous hidden state ht-1 and the current input xt on the candidate value; [ht-1,xt]: the concatenation of the previous hidden state ht-1 and the current input xt, as the input of the candidate value; bC: bias term of candidate value, used to adjust the output of candidate value; Unit status update formula: Among them: Ct represents the cell state at time step t, which is the memory of the LSTM cell; ft represents the output of the forget gate, which determines how to forget or retain the previous unit state Ct-1; Ct-1 represents the cell state at the previous time step; It represents the output of the input gate, which determines how to add the candidate value C~t to the cell state Output gate formula: o t =σ(W o ·[h t-1 ,x t ]+b o ); Where: ot represents the output of the output gate at time step t; Wo represents the weight matrix of the output gate, which determines the influence of the previous hidden state ht-1 and the current input xt on the output gate; [ht-1,xt] represents the concatenation of the previous hidden state ht-1 and the current input xt, which serves as the input of the output gate; bo represents the bias term of the output gate, which is used to adjust the output of the output gate Hidden state formula: Where: ht: hidden state at time step t, which is the output of the LSTM unit; ot represents the output of the output gate, which determines which parts of the cell state Ct will be output; Tanh(Ct) means that the cell state Ct is processed by the hyperbolic tangent function and the value is compressed to the interval [-1,1] S3-4. Initialize the XGBoost model parameters, including the number of trees, tree depth, and learning rate; S3-5. Model training: Use the training set to train the DNN, RNN and XGBoost models respectively; DNN training: Update weights through backpropagation and gradient descent; RNN training: Also use backpropagation and consider the gradient of the time series; XGBoost training: Use the gradient boosting framework to build a decision tree and optimize the objective function; S4. Model fusion: Use the DNN and RNN models to perform feature conversion on the training set data, obtain the output of the hidden layer as a new feature representation, combine the output features of the DNN and RNN models with the original features to form a new feature set, and then use the fused feature set to train the XGBoost model; the feature fusion formula is: X_{fusion}=[X_{DNN},X_{RNN},X_{original}] S5. Model application: transform the features of the samples to be evaluated through DNN and RNN models; input the transformed features into the XGBoost model for final prediction.

7. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The achievement transformation module is the core execution module of the intelligent evaluation and transformation management system for scientific and technological achievements. It is responsible for bringing the evaluated scientific and technological achievements to the market and realizing their commercial value. The specific content and operation methods of this module include: first, analyzing the comprehensive evaluation results of scientific and technological achievements to determine their transformation potential and appropriate transformation path; second, the module will automatically match potential cooperative enterprises or investment institutions and recommend the best partners through algorithms; then, the module provides a complete set of transformation process management tools, including project planning, fund raising, team building, and market promotion.

8. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The expert review module first selects experts with corresponding professional background and review experience from the expert database according to the fields and characteristics of the scientific and technological achievements; secondly, the module automatically sends review invitations to the selected experts and provides them with an online review platform to facilitate the experts to conduct reviews anytime and anywhere; next, the experts conduct a comprehensive evaluation of the originality, practicality and market prospects of the scientific and technological achievements through the platform, and give specific scores and suggestions; the module collects the review results of all experts and obtains the final review report through algorithm analysis; finally, the module will feedback the review report to the submitter of the scientific and technological achievements, and provide a reference for the management's decision-making.

9. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The application layer module operation method includes: defining various business processes in the system, including the results evaluation process and the transformation suggestion generation process; calling the API interface provided by the service layer to perform specific business operations, including generating an evaluation report and providing transformation path suggestions; processing the user's business request and converting the processing results of the service layer into a format that the user can understand.

10. The intelligent evaluation and transformation management system for scientific and technological achievements according to claim 1, characterized in that: The user interface layer module displays the services provided by the application layer to the user in the form of an intuitive graphical interface; the operation method includes: designing a friendly user interface so that users can easily operate and obtain information; realizing the input and output of user operations, including form submission, button clicks, and information display; collecting user input and requests, and passing them to the application layer for processing; receiving the processing results of the application layer, and displaying them to the user in the form of charts, lists, and reports;.

Images