Enterprise department and wound index prediction method and device, computer equipment and storage medium

Abstract

The invention discloses an enterprise department and wound index prediction method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring basic information of a to-be-predicted enterprise; performing feature engineering processing on the to-be-predicted enterprise basic information to obtain a feature processing result; inputting the feature processing result into a prediction model to carry out medical wound index prediction so as to obtain an enterprise medical wound index; wherein the prediction model is obtained by collecting public data of enterprise department and wound activities and basic information of corresponding enterprises, building a sample set and training a basic network of a MoE architecture by using the sample set; and outputting the enterprise department and wound index. By implementing the method, the technical problems existing in the aspects of comprehensiveness, timeliness, deep analysis ability and the like of an existing enterprise innovation ability evaluation method can be solved, comprehensive, dynamic and deep evaluation of the enterprise innovation ability is achieved, and scientific decision support is provided.

Description

Technical Field

[0001] The present invention relates to an exponential prediction method, and more specifically to an enterprise science and innovation index prediction method, device, computer device, and storage medium. Background Art

[0002] Currently, for the evaluation of an enterprise's innovation ability, most methods in the market rely on traditional scoring systems or static financial and basic information analysis. These traditional methods mainly focus on summarizing the historical data of the enterprise and considering it from a single dimension. For example, the innovation level of a company is measured by the proportion of financial investment or the number of patents held. However, this method has significant limitations. Specifically, traditional evaluation methods tend to focus on certain specific indicators, such as the number of patent applications or R & D capital investment, while ignoring other equally crucial factors, including but not limited to technology transformation efficiency, professional certifications obtained by the enterprise, and the results of innovation competitions participated in. This one-sided perspective limits the true reflection of the overall scientific and technological innovation strength of the enterprise. Existing evaluation means usually collect data based on fixed time points, which means that once the evaluation is completed, the results will no longer change with the development of the enterprise's innovation activities. Therefore, such evaluations cannot timely capture and reflect the latest progress or new achievements of the enterprise in scientific and technological innovation, resulting in information lag. In the process of feature extraction and model establishment, existing methods often adopt relatively simple linear analysis and fail to fully explore the complex relationships among the multi-dimensional innovation data of the enterprise. This not only weakens the accuracy of the prediction model but also limits the in-depth understanding of the factors affecting the enterprise's innovation ability, reducing the scientific nature and reliability of the evaluation conclusion.

[0003] Therefore, it is necessary to design a new method to solve the technical problems existing in the existing enterprise innovation ability evaluation methods in terms of comprehensiveness, timeliness, and in-depth analysis ability, achieve a comprehensive, dynamic, and in-depth evaluation of the enterprise's science and innovation ability, and provide scientific decision-making support. Summary of the Invention

[0004] The purpose of the present invention is to overcome the defects of the prior art and provide an enterprise science and innovation index prediction method, device, computer device, and storage medium.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: An enterprise science and innovation index prediction method, including:

[0006] Obtain the basic information of the enterprise to be predicted;

[0007] Perform feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result;

[0008] Input the feature processing result into a prediction model for science and technology innovation index prediction to obtain the enterprise science and technology innovation index; wherein, the prediction model is obtained by collecting the public data of the enterprise's science and technology innovation activities and the basic information of the corresponding enterprise, building a sample set, and training the basic network of the MoE architecture using the sample set;

[0009] Output the enterprise science and technology innovation index.

[0010] A further technical solution thereof is: the feature engineering processing of the basic information of the enterprise to be predicted to obtain a feature processing result includes:

[0011] Extract features related to science and technology innovation ability from the basic information of the enterprise to be predicted to obtain initial features;

[0012] Use XGBoost to evaluate the importance of the initial features, perform normalization processing, and filter out the normalized features with a cumulative importance less than a threshold to construct a feature table;

[0013] Process the feature table to obtain a feature processing result.

[0014] A further technical solution thereof is: the processing of the feature table to obtain a feature processing result includes:

[0015] Perform standardization of numerical features and encoding conversion of categorical features on the feature table to obtain a feature processing result.

[0016] A further technical solution thereof is: the prediction model is obtained by collecting the public data of the enterprise's science and technology innovation activities and the basic information of the corresponding enterprise, building a sample set, and training the basic network of the MoE architecture using the sample set, including:

[0017] Obtain the public data of the enterprise's science and technology innovation activities and the basic information of the corresponding enterprise to build a resource library containing all the basic information and science and technology innovation activity data of the enterprises;

[0018] Select enterprises with poor and good science and technology innovation capabilities as samples from the database, label the corresponding basic information of the selected enterprises to form a training set, and construct a full-scale sample set according to the training set;

[0019] Perform feature engineering processing on the basic information in the full-scale sample set to obtain a processing result;

[0020] Construct a basic network of the MoE architecture;

[0021] Define an activation function and a loss function;

[0022] Train the basic network using the processing result, and determine the prediction model in combination with the activation function and the loss function.

[0023] Its further technical solution is: the basic network includes an input layer, a feature fusion layer, a gating layer, multiple expert network layers, a fusion layer, and an output layer. Among them, each expert network is constructed based on an MLP, and tasks are dynamically assigned to different expert networks through a gating mechanism.

[0024] Its further technical solution is: the definition of the activation function and the loss function includes:

[0025] Use ReLU as the activation function, MSE as the loss function, and adopt the gradient descent method for optimization.

[0026] Its further technical solution is: the process of training the basic network using the processing result and determining the prediction model in combination with the activation function and the loss function includes:

[0027] Train the basic network using the processing result, use the gradient descent algorithm to iteratively update the model weights in combination with the activation function and the loss function, and at the same time adopt the cross-validation method to prevent overfitting and adjust the hyperparameters to determine the prediction model.

[0028] The present invention also provides an enterprise science and technology innovation index prediction device, including:

[0029] An information acquisition unit for acquiring the basic information of the enterprise to be predicted;

[0030] A feature engineering unit for performing feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result;

[0031] A prediction unit for inputting the feature processing result into the prediction model to perform science and technology innovation index prediction to obtain the enterprise science and technology innovation index; wherein, the prediction model is obtained by collecting the public data of the enterprise's science and technology innovation activities and the basic information of the corresponding enterprise, building a sample set, and training the basic network of the MoE architecture using the sample set;

[0032] An output unit for outputting the enterprise science and technology innovation index.

[0033] The present invention also provides a computer device, which includes a memory and a processor. A computer program is stored on the memory, and when the processor executes the computer program, the above method is implemented.

[0034] The present invention also provides a storage medium, which stores a computer program, and when the computer program is executed by a processor, the above method is implemented.

[0035] The beneficial effects of the present invention compared with the prior art are as follows: By obtaining detailed basic information of the enterprise to be predicted, the present invention ensures the comprehensiveness and diversity of data coverage, which not only includes traditional financial and operational indicators, but also covers specific details of innovation activities such as patent applications and scientific research investments, thus constructing a more comprehensive enterprise portrait; Through efficient feature engineering processing of the collected enterprise basic information, key features can be quickly extracted and feature processing results can be generated. This instant data processing mechanism ensures that the evaluation model can respond based on the latest enterprise information, improving the timeliness of the evaluation results; Input the data optimized by feature engineering into a specially trained MoE architecture prediction model. This model is trained using a sample set built from a large amount of public data on enterprise scientific and technological innovation activities and corresponding basic information, and has strong pattern recognition and prediction capabilities, capable of deeply exploring the enterprise innovation potential and development trends hidden behind the data; The finally output enterprise science and technology innovation index provides an intuitive and guiding quantitative evaluation standard for enterprise management. This index not only reflects the current enterprise innovation ability status, but also helps the enterprise identify its competitive advantages and improvement spaces, providing solid data support for strategic planning and resource allocation, and realizing the effective transformation from data to wisdom.

[0036] The following further describes the present invention in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts.

[0038] Figure 1 It is a schematic diagram of the application scenario of the enterprise science and technology innovation index prediction method provided by the embodiment of the present invention;

[0039] Figure 2 It is a schematic diagram of the process of the enterprise science and technology innovation index prediction method provided by the embodiment of the present invention;

[0040] Figure 3 It is a schematic diagram of the sub-process of the enterprise science and technology innovation index prediction method provided by the embodiment of the present invention Figure 1 ;

[0041] Figure 4 It is a schematic diagram of the sub-process of the enterprise science and technology innovation index prediction method provided by the embodiment of the present invention Figure 2 ;

[0042] Figure 5 It is a schematic diagram of the working process of the gating layer and multiple expert network layers provided by the embodiment of the present invention;

[0043] Figure 6 Schematic diagram of the working process of the output layer provided by the embodiment of the present invention;

[0044] Figure 7 Schematic block diagram of the enterprise science and technology innovation index prediction device provided by the embodiment of the present invention;

[0045] Figure 8 Schematic block diagram of the computer device provided by the embodiment of the present invention. Detailed implementation manners

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0047] It should be understood that when used in this specification and the appended claims, the terms "comprises" and "comprising" indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0048] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in this specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0049] It should be further understood that the term " / and / " as used in this specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0050] Please refer to Figure 1 and Figure 2 , Figure 1 Schematic diagram of the application scenario of the enterprise science and technology innovation index prediction method provided by the embodiment of the present invention. Figure 2Schematic flowchart of the enterprise science and technology innovation index prediction method provided by the embodiment of the present invention. This enterprise science and technology innovation index prediction method is applied to a server. The server conducts data interaction with terminals. In order to improve the accuracy and comprehensiveness of the evaluation of enterprise science and technology innovation capabilities, this method integrates data from multiple dimensions such as enterprise basic information, patent data, scientific research capital investment, and transformation of scientific and technological achievements, ensuring that the innovation potential and actual performance of enterprises are captured from multiple perspectives. On this basis, a multi-task learning model based on the MoE (Mixture of Experts) architecture is introduced. Different types of enterprise science and technology innovation activities are assigned to specially designed expert networks for processing. Each expert network focuses on a specific task or field, such as patent analysis, evaluation of the return on scientific research investment, etc., so as to achieve a more refined and targeted evaluation of enterprise innovation capabilities. This approach effectively overcomes the problem that traditional single models are difficult to adapt to the requirements of multiple tasks simultaneously, and provides a more flexible and accurate evaluation solution. In addition, the method of this embodiment has strong real-time data update and dynamic prediction capabilities. By implementing an automated data collection and processing mechanism, a continuously updated science and technology innovation resource library can also be enabled. This not only ensures that the evaluation model always runs based on the latest enterprise activity information, but also enables the evaluation results to be adjusted in a timely manner as the enterprise develops and changes, enhancing the timeliness and accuracy of the prediction.

[0051] Figure 2 It is a schematic flowchart of the enterprise science and technology innovation index prediction method provided by the embodiment of the present invention. As Figure 2 shown, this method includes the following steps S110 to S140.

[0052] S110. Obtain the basic information of the enterprise to be predicted.

[0053] In this embodiment, the basic information of the enterprise to be predicted includes but is not limited to company name, registered capital, establishment time, business status, etc. These information can be obtained through public channels such as enterprise industrial and commercial registration information and annual reports.

[0054] Specifically, the company name is the basic identifier for identifying an enterprise and plays an important role in understanding the enterprise's brand positioning and market influence. By analyzing the company name, its industry field and possible business scope can be initially judged.

[0055] The registered capital reflects the financial strength of the enterprise at the time of establishment and the initial investment scale of the shareholders in the company. A higher registered capital usually means stronger financial support capabilities and greater operational flexibility, which has a direct impact on scientific research investment and innovation activities.

[0056] The establishment time of an enterprise can help determine the length of its market presence, and thus infer the experience and technological accumulation it has built up in a specific industry. Established enterprises may have more mature management and technological systems, while newly founded enterprises may be more innovative and adaptable to rapid changes.

[0057] The operating status refers to whether the enterprise is in normal operation, undergoing rectification, in bankruptcy liquidation, etc. at the current stage. This information helps to understand the current financial health and operational stability of the enterprise, and thus affects the assessment of its future innovation ability and development potential.

[0058] Other relevant information (obtained from annual reports and other public channels):

[0059] Financial performance: including key financial indicators such as operating revenue, net profit, total assets, etc., which are used to measure the economic strength and profitability of the enterprise.

[0060] Management information: such as the composition of the board of directors, the senior management team and their background experiences, which is closely related to the strategic decision-making direction and execution ability of the enterprise.

[0061] Product or service lines: Describe the main types of products and services provided by the enterprise, as well as their competitive positions in the market.

[0062] Ratio of R & D expenditure: The proportion of annual R & D investment in total revenue, which reflects the degree of emphasis on technology R & D and the level of continuous investment of the enterprise.

[0063] Intellectual property situation: Patent numbers, trademark registrations, copyrights, etc., which demonstrate the technological innovation achievements and technological barriers of the enterprise.

[0064] Partnerships: Cooperative relationships with academic institutions, research institutes and other enterprises, which indicate the enterprise's ability to utilize external resources and collaborative innovation.

[0065] Fulfillment of social responsibilities: The practices of the enterprise in aspects such as environmental protection and employee welfare, which reflect its considerations for long-term sustainable development.

[0066] The above information together constitutes the basic framework for a comprehensive and in-depth analysis of the enterprise, providing solid data support for accurately evaluating its scientific and technological innovation ability and future development potential. By integrating multi-source heterogeneous data and combining advanced data analysis techniques and models, a more objective and scientific evaluation result of the enterprise's scientific and technological innovation ability can be provided.

[0067] S120. Perform feature engineering processing on the basic information of the enterprise to be predicted to obtain the feature processing result.

[0068] In this embodiment, the feature processing result refers to the optimized dataset that has undergone a series of data preprocessing and feature engineering steps and is ready for model training.

[0069] In one embodiment, refer to Figure 3 , the above step S120 may include steps S121 to S123.

[0070] S121. Extract features related to scientific and technological innovation capabilities from the basic information of the enterprise to be predicted to obtain initial features.

[0071] In this embodiment, the initial features include but are not limited to the following:

[0072] Number of patents: Count the number of patent applications, authorizations, and valid patents of the enterprise.

[0073] Technological innovation awards: Record the number of technological innovation awards obtained by the enterprise.

[0074] Investment in scientific research funds: Based on financial statements or special scientific research fund records, count the annual investment amount in scientific research.

[0075] These features are extracted from multiple data sources (such as enterprise dimension tables, patent cleaning tables, patent analysis tables, high-tech enterprise qualification tables, etc.) and organized into an initial feature set, providing a basis for subsequent feature selection and optimization.

[0076] S122. Use XGBoost to evaluate the importance of the initial features, perform normalization processing, and filter out the normalized features with cumulative importance less than the threshold to construct a feature table.

[0077] In this embodiment, the machine learning algorithm XGBoost is used to evaluate the influence degree of each feature on the target variable (i.e., the scientific and technological innovation ability score). By calculating the importance scores of each feature and normalizing them, the sum of the importance of all features is equal to 1. Then, sort them according to importance and set a threshold (for example, 0.8), and only retain those features whose cumulative importance does not exceed the threshold. This can remove redundant or irrelevant features and improve the efficiency and generalization ability of the model.

[0078] Specifically, the feature table is a table constructed based on the selected features. In addition to the necessary features, the feature table also includes basic information fields such as the unified social credit code of the enterprise and the company name. The specific construction method is to merge various features (such as the number of patents, technological innovation awards, scientific research funds invested, etc.) by enterprise to form a complete feature dataset. The basic information of the enterprise, scientific and technological innovation activity data, scientific research investment, etc. will be uniformly mapped to a table. Specifically, as shown in Table 1, the enterprise dimension table, which contains the basic information of the enterprise and is used to uniquely identify each enterprise and its related attributes. These attributes will serve as foreign keys in other tables to help associate the patents and qualification information of the enterprise.

[0079] Table 1. Enterprise Dimension Table

[0080]

[0081]

[0082] As shown in Table 2, the patent cleaning table stores the information of all patents of the enterprise, including the application date, legal status, inventors, IPC classification number, etc. It provides the original patent information and provides data support for subsequent feature extraction.

[0083] Table 2. Patent Cleaning Table

[0084]

[0085]

[0086] As shown in Table 3, the patent analysis table is the analysis data based on the patent cleaning table, mainly used to count the number of patents and patent features in different time periods. For example, the number of new patents, valid patents, invention patents, etc., which can provide the historical performance of various indicators for the model.

[0087] Table 3. Patent Analysis Table

[0088]

[0089]

[0090] As shown in Table 4, the science and technology enterprise qualification table contains the science and technology qualification information obtained by the enterprise, helps to identify which enterprises have qualifications such as high-tech recognition, and the effective status of these qualifications. It will be combined with other features as an important indicator of the enterprise's scientific and technological capabilities.

[0091] Table 4. Science and Technology Enterprise Qualification Table

[0092]

[0093]

[0094] As shown in Table 5, in the Sci-Tech Innovation Index prediction model, the relevant characteristics of an enterprise are extracted and stored in a characteristic table named the Sci-Tech Innovation Index Characteristic Table. The fields of this table are sourced from multiple data sources, such as the enterprise dimension table, the patent cleaning table, the patent analysis table, the high-tech enterprise qualification table, etc. This table is used to comprehensively describe the enterprise's sci-tech innovation activities.

[0095] Table 5. Sci-Tech Innovation Index Characteristic Table

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104] The fields of these characteristic tables cover multi-dimensional characteristics of an enterprise in terms of scientific and technological innovation, which can help the model more accurately evaluate the enterprise's innovation ability. These characteristic data will go through the steps of feature processing and standardization and finally be used as the input of the model to predict the enterprise's Sci-Tech Innovation Index.

[0105] S123. Process the said characteristic table to obtain a feature processing result.

[0106] In this embodiment, numerical feature standardization and categorical feature encoding conversion are performed on the said characteristic table to obtain a feature processing result.

[0107] For continuous or numerical features, such as registered capital, scientific research funds investment, etc., apply standardization methods (such as Z-score standardization) to make data of different magnitudes comparable.

[0108] For non-numerical or categorical features, such as enterprise type, industry code, etc., use methods such as One-Hot Encoding or Label Encoding to convert them into numerical forms for easy understanding and processing by the model.

[0109] If there are features related to the time dimension, it is necessary to consider how to reasonably represent the time attributes of these features, such as calculating the change trend in the past period through window functions.

[0110] Check and handle any possible missing value problems to ensure data integrity.

[0111] Identify and appropriately handle outliers to prevent them from having a negative impact on model training.

[0112] In summary, the feature processing results are a set of high-quality features that have been carefully selected, optimized, and standardized. It not only reflects the basic information of the enterprise and the key indicators of its scientific and technological innovation activities, but also lays a solid foundation for the subsequent establishment of an accurate scientific and technological innovation ability prediction model.

[0113] S130. Input the feature processing results into the prediction model for scientific and technological innovation index prediction to obtain the enterprise's scientific and technological innovation index; wherein, the prediction model is obtained by collecting the public data of the enterprise's scientific and technological innovation activities and the basic information of the corresponding enterprise, building a sample set, and training the basic network of the MoE architecture using the sample set.

[0114] In this embodiment, the enterprise's scientific and technological innovation index refers to the enterprise's scientific and technological innovation ability.

[0115] In one embodiment, please refer to Figure 4 , the prediction model is obtained by collecting the public data of the enterprise's scientific and technological innovation activities and the basic information of the corresponding enterprise, building a sample set, and training the basic network of the MoE architecture using the sample set, and may include steps S131 to S136.

[0116] S131. Obtain the public data of the enterprise's scientific and technological innovation activities and the basic information of the corresponding enterprise to build a resource library containing all the basic information of the enterprise and the data of scientific and technological innovation activities.

[0117] In this embodiment, collect the scientific and technological innovation activity data and its basic information of the enterprise from public channels (such as designated websites, patent databases, industry association reports, etc.). These data should cover aspects such as the enterprise's patent applications, authorizations, and the number of valid patents, technological innovation awards, and scientific research funds invested. Specifically, as shown in Table 6.

[0118] Table 6. Explanation of Scientific and Technological Innovation Activity Data

[0119]

[0120] Organize and store the collected data in a structured database to ensure data integrity and consistency. This resource library not only contains the static information of the enterprise (such as unified social credit code, company name, etc.), but also records the dynamic scientific and technological innovation activity indicators.

[0121] S132. Select enterprises with poor and good technological innovation capabilities from the database as samples, label the basic information corresponding to the selected enterprises to form a training set, and construct a full sample set according to the training set.

[0122] In this embodiment, based on historical performance or industry standards, some enterprises that are outstanding (i.e., high innovation ability) and relatively weak (i.e., low innovation ability) in technological innovation are selected as positive and negative samples.

[0123] Assign corresponding labels to each selected sample. For example, "high innovation ability enterprise", "medium innovation ability enterprise" or "low innovation ability enterprise". This step is crucial for supervised learning as it provides the target variable, enabling the model to learn how to distinguish different levels of innovation.

[0124] Integrate the above positive and negative samples to form a training set that comprehensively reflects the differences in enterprise innovation capabilities for subsequent model training.

[0125] Specifically, select enterprises with relatively low technological innovation capabilities from the above technological innovation activity data as bad samples. The specific criteria are: enterprises whose performance in technological innovation activities such as patent application quantity, scientific research investment, and technology conversion rate is lower than the industry average level. In the past 12 months of their technological innovation activities, these enterprises have a patent application quantity lower than the industry median, a lower proportion of scientific research funds invested in revenue, a lower technology achievement conversion rate than the industry average, and have not won significant technological innovation awards, etc. These samples will be used as negative samples during training.

[0126] Select enterprises with good performance in technological innovation activities. The specific criteria are: indicators in aspects such as patent application, scientific research investment, and technology conversion are better than the industry average level. For example: the patent application quantity is higher than the industry average, the proportion of scientific research funds invested in revenue is high, having obtained high-tech enterprise certification, or having a high technology achievement conversion rate, etc. Ensure that the samples are representative and balanced. The good samples and bad samples will be labeled to form a training set.

[0127] S133. Perform feature engineering processing on the basic information in the full sample set to obtain a processing result.

[0128] In this embodiment, the above feature engineering processing can refer to the feature engineering processing in step S120 and will not be elaborated here. The purpose is to generate high-quality feature vectors as the input for the next model training. The specific operations include but are not limited to standardization of numerical features, encoding of categorical features, adjustment of time series features, etc.

[0129] S134. Construct a basic network with a MoE architecture.

[0130] In this embodiment, please refer to Figure 5 and Figure 6 , the basic network includes an input layer, a feature fusion layer, a gating layer, multiple expert network layers, a fusion layer, and an output layer. Among them, each expert network is constructed based on an MLP, and tasks are dynamically assigned to different expert networks through a gating mechanism.

[0131] In this embodiment, according to the design concept of the MoE architecture, a deep neural network composed of an input layer, a feature fusion layer, a gating layer, multiple expert network layers, a fusion layer, and an output layer is constructed. Each expert network is implemented based on an MLP (Multi-Layer Perceptron) and focuses on processing specific scientific and technological innovation tasks (such as patent application prediction, scientific research investment prediction, etc.). The gating mechanism is responsible for dynamically assigning tasks to the most suitable expert network according to the input features.

[0132] S135. Define the activation function and the loss function.

[0133] In this embodiment, ReLU is used as the activation function, MSE is used as the loss function, and the gradient descent method is used for optimization.

[0134] In this embodiment, ReLU is selected as the activation function for the hidden layer to introduce non-linearity and help the model better capture complex patterns. The mean squared error (MSE) is used as the loss function to measure the gap between the model prediction value and the true value. At the same time, in the multi-task learning scenario, the total loss can be calculated by weighted averaging.

[0135] S136. Use the processing result to train the basic network, and determine the prediction model in combination with the activation function and the loss function.

[0136] In this embodiment, the basic network is trained using the processing result. The gradient descent algorithm is used to iteratively update the model weights in combination with the activation function and the loss function. At the same time, the cross-validation method is adopted to prevent overfitting and adjust the hyperparameters to determine the prediction model.

[0137] In this embodiment, the sample set processed by feature engineering is used to train the basic network of the MoE architecture. During this process, the gradient descent algorithm (such as the Adam optimizer) is applied to iteratively update the model weights to ensure that the model parameters gradually converge to the optimal solution.

[0138] Adopt a cross-validation strategy (such as K-fold cross-validation), evaluate the model performance on the validation set, and adjust the hyperparameters in a timely manner to avoid the model overfitting the training data.

[0139] When the performance of the model on the validation set is stable and meets expectations, it is determined as the final prediction model. At this time, the model already has the ability to accurately evaluate the technological innovation ability of enterprises and can be applied to actual business scenarios.

[0140] In this embodiment, the overall network architecture of the entire prediction model adopts the Mixture of Experts (MoE) architecture. This architecture dynamically assigns tasks to different expert networks for processing through a gating mechanism, thereby replacing the traditional single-task learning model. Each expert network is constructed based on a multi-layer perceptron (MLP). They share the same input features, but according to different selections of the gating mechanism, they will be optimized for specific tasks. This design allows the model to adapt to various scientific and technological innovation tasks, such as patent applications, technological innovation, etc.

[0141] The outputs of each expert network are integrated in the fusion layer to form a comprehensive enterprise scientific and technological innovation index as the final output, reflecting the overall performance of the enterprise in scientific and technological innovation.

[0142] The input layer includes:

[0143] Numerical feature input: including the registered capital of the enterprise, annual scientific research funds input, the number of inventors of invention patents, the number of valid patents, the proportion of invalidated patents, the average remaining validity period of invention patents, the average holding period of non-invention patents, as well as the number of patent pledges and the number of approved trademarks, etc.

[0144] Text feature input: involving descriptive text data such as enterprise industry descriptions, patent IPC classification types, and science and technology enterprise type labels. These data are converted into a machine-readable form through natural language processing technology.

[0145] After preprocessing the above numerical and text features, the feature fusion layer generates a unified feature vector through feature concatenation or other fusion methods as the input for subsequent steps.

[0146] The feature vector enters the gating layer after being processed by a multi-layer fully connected network. The gating mechanism dynamically selects the most suitable expert network to handle specific scientific and technological innovation tasks according to the input features. For example:

[0147] E1: Predict the number of patent applications and their year-on-year growth rate;

[0148] E2: Evaluate the investment in scientific research funds;

[0149] E3: Predict the growth of the number of valid patents;

[0150] E4: Analyze the type of science and technology enterprises;

[0151] E5: Predict the average remaining validity period of invention patents;

[0152] E6: Analyze the status of enterprise patent pledge;

[0153] E7: Analyze the proportion of the number of times a patent is invalidated;

[0154] E8: Analyze industry classification;

[0155] E9: Analyze the innovation activities of enterprises in strategic emerging industries.

[0156] Each expert network uses an MLP for non - linear mapping internally, focuses on processing specific tasks, and finally integrates the outputs of all expert networks through a fusion layer to generate a unified enterprise science and innovation index.

[0157] The fusion layer aggregates the information learned by all expert networks and provides a more comprehensive evaluation of the enterprise's innovation ability.

[0158] In the output layer, combining the fusion outputs of all expert networks, the enterprise's science and innovation index is calculated through a Softmax layer or other appropriate functions. This index is used to classify the enterprise's innovation ability and provide a basis for decisions such as investment and cooperation.

[0159] Use ReLU as the activation function, especially for the hidden layer, to introduce non - linearity and help the model better learn complex features.

[0160] Select the mean squared error (MSE) as the loss function. In the context of multi - task learning, MSE is used to measure the prediction error of each task and is weighted and combined in the final fusion layer. The gradient descent method (such as the Adam optimizer) may be used during the optimization process.

[0161] Before training, the sample data needs to be cleaned, missing values processed, and normalized to ensure data quality.

[0162] Use the gradient descent algorithm (such as the Adam optimizer) to iteratively update the model weights, and adopt the cross - validation method to prevent overfitting. At the same time, adjust the hyperparameters on the validation set.

[0163] The performance evaluation on the test set mainly focuses on MSE and prediction stability. If the prediction error is large, the model can be optimized by adjusting the network structure, increasing the training data, or adjusting the learning rate, etc.

[0164] Real - time obtain the latest science and innovation data of enterprises through the API interface and convert it into a format suitable for model input through a pre - processing module.

[0165] Input the processed data into the trained model to obtain a science and innovation index reflecting the enterprise's performance in aspects such as innovation, scientific research investment, and technology transformation.

[0166] Classify the enterprise innovation ability according to the science and technology innovation index, providing a scientific basis for decision-makers.

[0167] As new data continuously flows in, the system retrains the model regularly to maintain its accuracy and real-time performance. Meanwhile, it optimizes the model prediction accuracy according to actual feedback, including but not limited to adding new features, improving the model structure, or adjusting the training strategy.

[0168] S140. Output the enterprise science and technology innovation index.

[0169] In this embodiment, the enterprise science and technology innovation index is output to the terminal for display.

[0170] In addition, traditional linear regression or random forest models can be considered for enterprise innovation ability assessment to reduce system complexity. However, this alternative is weaker in model accuracy and multitasking ability and cannot achieve the comprehensive performance of the present patented technology. The dimension of data collection can be reduced, only retaining key indicators (such as the number of patents and investment in scientific research funds), to reduce the complexity of data collection and processing. Although this method can reduce the computational cost of the system, it will sacrifice the comprehensiveness and accuracy of the assessment.

[0171] The above-mentioned enterprise science and technology innovation index prediction method ensures the wide coverage and diversity of data by obtaining detailed basic information of the enterprise to be predicted, which includes not only traditional financial and operation indicators but also specific details of innovation activities such as patent applications and scientific research investments, thus constructing a more comprehensive enterprise portrait; through efficient feature engineering processing of the collected enterprise basic information, key features can be quickly extracted and the feature processing results can be generated. This immediate data processing mechanism ensures that the evaluation model can respond based on the latest enterprise information, improving the timeliness of the evaluation results; the data optimized by feature engineering is input into a specially trained MoE architecture prediction model. This model is trained using a large amount of publicly available data on enterprise science and technology innovation activities and corresponding basic information to build a sample set, and has strong pattern recognition and prediction capabilities, capable of deeply exploring the innovation potential and development trends of enterprises hidden behind the data; the finally output enterprise science and technology innovation index provides an intuitive and guiding quantitative evaluation standard for enterprise management. This index not only reflects the current enterprise innovation ability status but also helps enterprises identify their competitive advantages and improvement spaces, providing solid data support for strategic planning and resource allocation, and realizing the effective transformation from data to wisdom.

[0172] Figure 7 It is a schematic block diagram of an enterprise science and technology innovation index prediction device 300 provided by an embodiment of the present invention. As Figure 7As shown, corresponding to the above enterprise scientific and technological innovation index prediction method, the present invention also provides an enterprise scientific and technological innovation index prediction device 300. The enterprise scientific and technological innovation index prediction device 300 includes units for executing the above enterprise scientific and technological innovation index prediction method, and the device can be configured in a server. Specifically, please refer to Figure 7 , the enterprise scientific and technological innovation index prediction device 300 includes an information acquisition unit 301, a feature engineering unit 302, a prediction unit 303, and an output unit 304.

[0173] The information acquisition unit 301 is used to acquire the basic information of the enterprise to be predicted; the feature engineering unit 302 is used to perform feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result; the prediction unit 303 is used to input the feature processing result into a prediction model for scientific and technological innovation index prediction to obtain an enterprise scientific and technological innovation index; wherein, the prediction model is obtained by collecting the public data of the enterprise scientific and technological innovation activities and the basic information of the corresponding enterprise, building a sample set, and using the sample set to train the basic network of the MoE architecture; the output unit 304 is used to output the enterprise scientific and technological innovation index.

[0174] In one embodiment, the feature engineering unit 302 includes:

[0175] An extraction subunit, configured to extract features related to scientific and technological innovation capabilities from the basic information of the enterprise to be predicted to obtain initial features; an evaluation subunit, configured to use XGBoost to evaluate the importance of the initial features, perform normalization processing, and filter out the normalized features with a cumulative importance less than a threshold to construct a feature table; a processing subunit, configured to process the feature table to obtain a feature processing result.

[0176] In one embodiment, the processing subunit is configured to perform standardization of numerical features and encoding conversion of categorical features on the feature table to obtain a feature processing result.

[0177] In one embodiment, the device further includes a training unit, configured to:

[0178] Acquire the public data of the enterprise scientific and technological innovation activities and the basic information of the corresponding enterprise to build a resource library containing all the basic information of the enterprise and the scientific and technological innovation activity data; select enterprises with poor scientific and technological innovation capabilities and good scientific and technological innovation capabilities from the database as samples, and label the corresponding basic information of the selected enterprises to form a training set, and construct a full-scale sample set according to the training set; perform feature engineering processing on the basic information in the full-scale sample set to obtain a processing result; construct a basic network of the MoE architecture; define an activation function and a loss function; use the processing result to train the basic network, and determine the prediction model in combination with the activation function and the loss function.

[0179] Specifically, ReLU is used as the activation function, MSE is used as the loss function, and the gradient descent method is adopted for optimization.

[0180] Train the basic network using the processing result, and use the gradient descent algorithm to iteratively update the model weights in combination with the activation function and the loss function. At the same time, adopt the cross-validation method to prevent overfitting and adjust the hyperparameters to determine the prediction model.

[0181] It should be noted that those skilled in the art can clearly understand that the specific implementation processes of the above-mentioned enterprise science and technology innovation index prediction device 300 and each unit can refer to the corresponding descriptions in the foregoing method embodiments. For the sake of convenience and conciseness of description, they will not be elaborated here.

[0182] The above-mentioned enterprise science and technology innovation index prediction device 300 can be implemented in the form of a computer program, and this computer program can run on a computer device as shown in Figure 8 shown.

[0183] Please refer to Figure 8 , Figure 8 which is a schematic block diagram of a computer device provided by an embodiment of the present application. The computer device 500 may be a server. Among them, the server may be an independent server or a server cluster composed of multiple servers.

[0184] Refer to Figure 8 , the computer device 500 includes a processor 502, a memory, and a network interface 505 connected through a system bus 501. Among them, the memory may include a non-volatile storage medium 503 and an internal memory 504.

[0185] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions. When the program instructions are executed, the processor 502 can be made to execute an enterprise science and technology innovation index prediction method.

[0186] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500.

[0187] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can be made to execute an enterprise science and technology innovation index prediction method.

[0188] The network interface 505 is used for network communication with other devices. Those skilled in the art can understand that Figure 8The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device 500 to which the solution of this application is applied. Specifically, the computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0189] Among them, the processor 502 is used to run the computer program 5032 stored in the memory to implement the following steps:

[0190] Obtain the basic information of the enterprise to be predicted; perform feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result; input the feature processing result into a prediction model for predicting the science and technology innovation index to obtain the enterprise science and technology innovation index; among them, the prediction model is obtained by collecting the public data of the enterprise's science and technology innovation activities and the basic information of the corresponding enterprise, building a sample set, and training the basic network of the MoE architecture using the sample set; output the enterprise science and technology innovation index.

[0191] In one embodiment, when the processor 502 implements the step of performing feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result, the following steps are specifically implemented:

[0192] Extract the features related to the science and technology innovation ability from the basic information of the enterprise to be predicted to obtain initial features; use XGBoost to evaluate the importance of the initial features, perform normalization processing, and screen out the normalized features with a cumulative importance less than a threshold to construct a feature table; process the feature table to obtain a feature processing result.

[0193] In one embodiment, when the processor 502 implements the step of processing the feature table to obtain a feature processing result, the following steps are specifically implemented:

[0194] Perform standardization of numerical features and encoding conversion of categorical features on the feature table to obtain a feature processing result.

[0195] In one embodiment, when the processor 502 implements the step that the prediction model is obtained by collecting the public data of the enterprise's science and technology innovation activities and the basic information of the corresponding enterprise, building a sample set, and training the basic network of the MoE architecture using the sample set, the following steps are specifically implemented:

[0196] Obtain the public data of the enterprise's scientific and technological innovation activities and the basic information of the corresponding enterprise to build a resource library containing all the basic information of the enterprise and the scientific and technological innovation activity data; select enterprises with poor scientific and technological innovation capabilities and good scientific and technological innovation capabilities from the said resource library as samples, and label the corresponding basic information of the selected enterprises to form a training set, and construct a full-scale sample set according to the training set; perform feature engineering processing on the basic information in the full-scale sample set to obtain a processing result; construct a basic network with a MoE architecture; define an activation function and a loss function; use the processing result to train the basic network, and determine the prediction model in combination with the activation function and the loss function.

[0197] Among them, the basic network includes an input layer, a feature fusion layer, a gating layer, multiple expert network layers, a fusion layer and an output layer. Among them, each expert network is constructed based on MLP, and tasks are dynamically assigned to different expert networks through a gating mechanism.

[0198] In one embodiment, when the processor 502 implements the step of defining the activation function and the loss function, the following steps are specifically implemented:

[0199] Use ReLU as the activation function, MSE as the loss function, and adopt the gradient descent method for optimization.

[0200] In one embodiment, when the processor 502 implements the step of using the processing result to train the basic network and determining the prediction model in combination with the activation function and the loss function, the following steps are specifically implemented:

[0201] Use the processing result to train the basic network, use the gradient descent algorithm to iteratively update the model weights in combination with the activation function and the loss function, and at the same time adopt the cross-validation method to prevent overfitting and adjust the hyperparameters to determine the prediction model.

[0202] It should be understood that in the embodiment of the present application, the processor 502 may be a central processing unit (Central Processing Unit, CPU), and this processor 502 may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application-specific integrated circuits (Application Specific Integrated Circuit, ASIC), field-programmable gate arrays (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc.

[0203] Those of ordinary skill in the art can understand that all or part of the processes in the methods of implementing the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the above method embodiments.

[0204] Therefore, the present invention also provides a storage medium. The storage medium can be a computer-readable storage medium. The storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the following steps:

[0205] Obtain the basic information of the enterprise to be predicted; perform feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result; input the feature processing result into a prediction model for predicting the science and technology innovation index to obtain the enterprise science and technology innovation index; wherein, the prediction model is obtained by collecting public data of the enterprise's science and technology innovation activities and the basic information of the corresponding enterprise, building a sample set, and training a basic network of the MoE architecture using the sample set; output the enterprise science and technology innovation index.

[0206] In one embodiment, when the processor executes the computer program to implement the step of performing feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result, the following steps are specifically implemented:

[0207] Extract features related to the science and technology innovation ability from the basic information of the enterprise to be predicted to obtain initial features; use XGBoost to evaluate the importance of the initial features, perform normalization processing, and screen out the normalized features with a cumulative importance less than a threshold to construct a feature table; process the feature table to obtain a feature processing result.

[0208] In one embodiment, when the processor executes the computer program to implement the step of processing the feature table to obtain a feature processing result, the following steps are specifically implemented:

[0209] Perform standardization of numerical features and encoding conversion of categorical features on the feature table to obtain a feature processing result.

[0210] In one embodiment, when the processor executes the computer program to implement the step that the prediction model is obtained by collecting public data of the enterprise's science and technology innovation activities and the basic information of the corresponding enterprise, building a sample set, and training a basic network of the MoE architecture using the sample set, the following steps are specifically implemented:

[0211] Obtain the public data of the enterprise's scientific and technological innovation activities and the basic information of the corresponding enterprise to build a resource library containing all the basic information of the enterprise and the scientific and technological innovation activity data; select the enterprises with poor scientific and technological innovation capabilities and good scientific and technological innovation capabilities from the said resource library as samples, and label the corresponding basic information of the selected enterprises to form a training set, and construct a full-scale sample set according to the training set; perform feature engineering processing on the basic information in the said full-scale sample set to obtain a processing result; construct a basic network with a MoE architecture; define an activation function and a loss function; use the said processing result to train the basic network, and determine the prediction model in combination with the said activation function and loss function.

[0212] Among them, the basic network includes an input layer, a feature fusion layer, a gating layer, multiple expert network layers, a fusion layer and an output layer. Among them, each expert network is constructed based on an MLP, and tasks are dynamically assigned to different expert networks through a gating mechanism.

[0213] In one embodiment, when the processor executes the computer program to implement the step of defining the activation function and the loss function, the following steps are specifically implemented:

[0214] Use ReLU as the activation function, MSE as the loss function, and adopt the gradient descent method for optimization.

[0215] In one embodiment, when the processor executes the computer program to implement the step of using the said processing result to train the basic network and determining the prediction model in combination with the said activation function and loss function, the following steps are specifically implemented:

[0216] Use the said processing result to train the basic network, use the gradient descent algorithm to iteratively update the model weights in combination with the said activation function and loss function, and at the same time adopt the cross-validation method to prevent overfitting and adjust the hyperparameters to determine the prediction model.

[0217] The storage medium can be various computer-readable storage media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disc that can store program codes.

[0218] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0219] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0220] The steps in the method embodiments of the present invention can be adjusted, combined, and deleted according to actual needs. The units in the device embodiments of the present invention can be combined, divided, and deleted according to actual needs. In addition, the functional units in each embodiment of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0221] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention.

[0222] As described above, the above are only the specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. The enterprise science and technology innovation index prediction method is characterized by: include: Obtain basic information of the enterprise to be predicted; Performing feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result; The feature processing result is input into the prediction model to predict the science and technology innovation index, so as to obtain the enterprise science and technology innovation index; wherein the prediction model is obtained by collecting public data of enterprise science and technology innovation activities and basic information of corresponding enterprises, building a sample set, and using the sample set to train the basic network of the MoE architecture; Output the enterprise's science and technology innovation index.

2. The enterprise science and technology innovation index prediction method according to claim 1 is characterized in that: The performing feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result includes: Extracting features related to scientific and technological innovation capabilities from the basic information of the enterprise to be predicted to obtain initial features; Using XGBoost to evaluate the importance of the initial features, perform normalization, and filter out normalized features whose cumulative importance is less than a threshold to construct a feature table; The feature table is processed to obtain a feature processing result.

3. The enterprise science and technology innovation index prediction method according to claim 2 is characterized in that: The processing of the feature table to obtain a feature processing result includes: The feature table is subjected to standardization of numerical features and encoding conversion of categorical features to obtain a feature processing result.

4. The enterprise science and technology innovation index prediction method according to claim 1 is characterized in that: The prediction model is obtained by collecting public data on enterprise scientific and technological innovation activities and basic information of the corresponding enterprises, building a sample set, and using the sample set to train the basic network of the MoE architecture, including: Obtain public data on corporate scientific and technological innovation activities and basic information of corresponding enterprises to build a resource library containing basic information of all enterprises and data on scientific and technological innovation activities; Selecting enterprises with poor scientific and technological innovation capabilities and enterprises with good scientific and technological innovation capabilities from the database as samples, labeling the basic information corresponding to the selected enterprises to form a training set, and constructing a full sample set based on the training set; Performing feature engineering processing on the basic information in the full sample set to obtain a processing result; Build the basic network of MoE architecture; Define activation function and loss function; The processing result is used to train the basic network, and the prediction model is determined in combination with the activation function and the loss function.

5. The enterprise science and technology innovation index prediction method according to claim 4 is characterized in that: The basic network includes an input layer, a feature fusion layer, a gating layer, multiple expert network layers, a fusion layer and an output layer, wherein each expert network is constructed based on MLP, and tasks are dynamically assigned to different expert networks through a gating mechanism.

6. The enterprise science and technology innovation index prediction method according to claim 4 is characterized in that: The definition of activation function and loss function includes: ReLU is used as the activation function, MSE is used as the loss function, and the gradient descent method is used for optimization.

7. The enterprise science and technology innovation index prediction method according to claim 4 is characterized in that: The using the processing result to train the basic network and determining the prediction model in combination with the activation function and the loss function includes: The processing results are used to train the basic network, and the gradient descent algorithm is used in combination with the activation function and the loss function to iteratively update the model weights. At the same time, a cross-validation method is used to prevent overfitting and adjust hyperparameters to determine the prediction model.

8. The enterprise science and technology innovation index prediction device is characterized by: include: An information acquisition unit, used to acquire basic information of the enterprise to be predicted; A feature engineering unit, used for performing feature engineering processing on the basic information of the enterprise to be predicted to obtain a feature processing result; A prediction unit is used to input the feature processing result into a prediction model to predict the science and technology innovation index, so as to obtain the enterprise science and technology innovation index; wherein the prediction model is constructed by collecting public data on the enterprise's science and technology innovation activities and basic information of the corresponding enterprise, and using the sample set to train the basic network of the MoE architecture; An output unit is used to output the enterprise science and technology innovation index.

9. A computer device, characterized in that: The computer device comprises a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images