Evaluation and prediction method for college ESI subject development situation

Abstract

The invention discloses an evaluation and prediction method for a college ESI subject development situation, belongs to the technical field of subject development situation evaluation and prediction, overcomes the defect of domestic college ESI subject development situation evaluation, and helps colleges to measure and calculate time nodes when ESI subjects enter one hundred percent of the global. And the possibility and the time period of continuing to stride towards one thousandth and one ten thousandth in the whole world. Efficient model training: the performance of the model can be improved when unbalanced data is processed by customizing an error evaluation function and a balance category weight. According to the multi-model architecture, twelve different model architectures are fused, and complex data requirements can be met. And automatic model storage and loading: the system automatically stores the optimal model and automatically loads the model during prediction, so that the model management process is simplified.

Description

Technical Field

[0001] The present invention belongs to the technical field of evaluation and prediction of discipline development trends, and specifically relates to an evaluation and prediction method for ESI discipline development trends in colleges and universities. Background Art

[0002] As mentioned in the prior art with patent publication number "CN110609875A", ESI is a quantitative analysis database constructed based on the literature data included in SCI and SSCI, and is currently widely used as a tool for scientific research performance evaluation at home and abroad. With the full launch of the construction of first-class universities and first-class disciplines in my country, how to effectively use ESI to evaluate the results of discipline construction and provide a basis for optimizing discipline layout and formulating development plans has far-reaching significance for promoting the construction of "Double First-Class", as follows: 1. The school hopes to use scientific analysis tools and big data prediction models to conduct forward-looking analysis of the future performance of its ESI disciplines. This includes specifically calculating the time point when the discipline enters the global 1%, and the possibility and time period of continuing to move towards the global 1 / 1000th and 1 / 10,000th. This method will help the school accurately identify the shortcomings of the discipline and develop a clear development path.

[0003] 2. The school needs to make accurate predictions on the development trend of its own ESI disciplines, which includes evaluating the growth of research results, changes in international rankings, and the improvement of overall academic influence in the disciplines in the next few years. This kind of prediction will not only help the school to clarify the discipline positioning, but also provide data support for the formulation of scientific development plans.

[0004] At present, a complete system has not been established for the evaluation of ESI disciplines, which makes it difficult for many universities to accurately grasp the development status and international competitiveness of their disciplines. It is unable to help schools accurately identify their discipline shortcomings and formulate clear development paths, nor can it provide data support for schools to clarify their discipline positioning and formulate scientific development plans. Summary of the invention

[0005] In order to solve the defects in the prior art, the present invention proposes an evaluation and prediction device and method for the development trend of ESI disciplines in colleges and universities, which effectively avoids the defects in the prior art that a complete system has not yet been established for the evaluation of ESI disciplines, many colleges and universities find it difficult to accurately grasp the development status and international competitiveness of disciplines, cannot help schools accurately locate discipline shortcomings and formulate clear development paths, and cannot provide data support for schools to clarify discipline positioning and formulate scientific development plans.

[0006] The present invention uses the following technical solutions.

[0007] An evaluation and prediction method for the development trend of ESI disciplines in colleges and universities, including: Step 1: Collect the 22 ESI subject ranking data of several domestic universities in the past two years to confirm the prediction object; Step 2: Divide the prediction object into training set and test set according to the set ratio, and customize the calculate function; Step 3: Construct a model for the development trend of ESI disciplines in universities; Step 4: Use the model for the development trend of ESI disciplines in colleges and universities to train the model; Step 5: Perform model prediction using the specified regression model.

[0008] Further, in step 1, the number of colleges and universities is more than 100 colleges and universities.

[0009] Furthermore, in step 1, the global 1% citations, global 1% citations, global 1% citations, global 1% citations and the university's own citations of 22 ESI disciplines are selected from the ESI subject ranking data of several domestic universities in the past two years as prediction objects.

[0010] Furthermore, in step 2, the ratio is set to 4:1, that is, the training set is 4 times the test set.

[0011] Furthermore, step 2 specifically includes: Convert the training set and test set as the target variable into classification format, that is, classify the target variable; The categories obtained by classifying the target variable are calculated to obtain their category weights to handle imbalanced data sets.

[0012] Furthermore, in step 2, the custom calculate function is used to calculate the error, that is, to calculate the error between the predicted value and the true value.

[0013] Furthermore, in step 2, the custom calculate function includes: Calculate the mean absolute error MAE, which is the average absolute value of the difference between the predicted value and the actual value; Calculate the variance of the mean absolute error MAE to measure the volatility of the variance; Returns the calculated mean absolute error (MAE) and its variance.

[0014] Furthermore, in step 3, the model for the development trend of ESI disciplines in colleges and universities is constructed, which specifically includes: Forecasting is performed using twelve regression methods, namely: Perform forecasts using a linear regression model; Predictions were performed using a ridge regression model; Predictions were performed using the lasso regression model; Forecasting was performed using the elastic net regression model; Predictions were performed using a support vector regression model; Perform predictions using a decision tree regression model; Predictions were performed using a random forest regression model; Perform predictions using a gradient boosting regression model; The prediction is performed using the AdaBoost regression model; Use XGBoost regression model to perform prediction; Use CatBoost regression model to perform prediction; Use a multi-layer perceptron regression model to perform predictions; The model built for the development trend of ESI disciplines in colleges and universities also includes: Use the twelve regression methods one by one to perform prediction and store the prediction results in a dictionary; Returns the prediction results obtained by performing prediction using the twelve regression methods described above.

[0015] Furthermore, in step 4, a method for model training using a model for the development trend of ESI disciplines in colleges and universities is specifically provided as follows: Send the training set to the model for the development trend of ESI disciplines in colleges and universities to train the model and obtain a trained model; Use the trained model to predict the next data point to get the predicted value; Use the custom calculate function to obtain the mean absolute error MAE of the training set and its variance; Returns the unnormalized forecast value, the mean absolute error (MAE), and the variance of the MAE.

[0016] Furthermore, in step 5, a scheme for executing model prediction through a specified regression model specifically includes: Get the newly prepared ESI subject ranking data; Use the trained model to predict the newly prepared ESI subject ranking data, and use the custom calculate function to obtain the mean absolute error MAE and the variance of the mean absolute error MAE of the newly prepared ESI subject ranking data; Returns the predicted value, mean absolute error (MAE) and the variance of the mean absolute error (MAE) of the newly prepared ESI subject ranking data.

[0017] An information management device for intelligent archives, comprising: Processing terminal. The modules running on the processing terminal include: Model partitioning module, model building module, model training module and model prediction module; The model partitioning module is used to divide the prediction object into training set and test set according to the set ratio, and also customize the calculate function; The model building module is used to build a model for the development trend of ESI disciplines in colleges and universities; The model training module is used to train models using models for the development trends of ESI disciplines in universities; The model prediction module is used to perform model prediction using a specified regression model.

[0018] The beneficial effects of the present invention are that, compared with the prior art, the technical effects of the present invention include: It improves the shortcomings of domestic assessment of the development trend of ESI disciplines in colleges and universities, and helps colleges and universities calculate the time node when ESI disciplines enter the global 1%, as well as the possibility and time period of continuing to move towards the global 1 / 1000 and 1 / 10,000. Efficient model training: By customizing the error evaluation function and balancing the category weights, the performance of the model can be improved when processing unbalanced data. Multi-model architecture: It integrates twelve different model architectures and can adapt to complex data requirements. Automated model saving and loading: The system automatically saves the best model and automatically loads the model during prediction, simplifying the model management process. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the evaluation and prediction method for the development trend of ESI disciplines in colleges and universities described in the present invention; Figure 2 It is a partial structural diagram of the evaluation and prediction device for the development trend of ESI disciplines in colleges and universities described in the present invention; Figure 3 It is a functional code schematic diagram of the evaluation and prediction method for the development trend of ESI disciplines in colleges and universities described in the present invention; Figure 4 It is a functional code schematic diagram of a custom calculate function of the evaluation and prediction method for the development trend of ESI disciplines in colleges and universities described in the present invention; Figure 5 It is a functional code schematic diagram of the model for constructing the evaluation and prediction method for the development trend of ESI disciplines in colleges and universities described in the present invention; Figure 6 It is a functional code schematic diagram of the method for evaluating and predicting the development trend of ESI disciplines in colleges and universities described in the present invention using a model for training the development trend of ESI disciplines in colleges and universities; Figure 7It is a functional code schematic diagram of the method for evaluating and predicting the development trend of ESI disciplines in colleges and universities described in the present invention, which performs model prediction through a specified regression model; Figure 8 It is the error coordinate diagram of the global 1% citation volume of the evaluation and prediction method for the development trend of ESI disciplines in colleges and universities described in the present invention; Fig. 9 It is the error coordinate diagram of the global one thousandth citation volume of the evaluation and prediction method for the development trend of ESI disciplines in colleges and universities described in the present invention; Fig.10 It is the error coordinate diagram of the global one ten-thousandth citation volume of the evaluation and prediction method for the development trend of ESI disciplines in colleges and universities described in the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely expressed in conjunction with the drawings in the embodiments of the present invention. The embodiments expressed in this application are only partial embodiments of the present invention, not all embodiments. According to the spirit of the present invention, other embodiments obtained by technicians in this field without creative work are all within the protection scope of the present invention.

[0021] like Figure 1 As shown, the evaluation and prediction method for the development trend of ESI disciplines in colleges and universities described in the present invention runs on a processing terminal and includes: Step 1: Collect the 22 ESI subject ranking data of several domestic universities in the past two years to confirm the prediction object; In a preferred but non-limiting embodiment of the present invention, in step 1, the number of colleges and universities is more than 100.

[0022] In a preferred but non-limiting implementation of the present invention, in step 1, the global 1% citation volume, global 1% citation volume, global 1% citation volume, global 1% citation volume and the university's own citation volume of 22 ESI disciplines are screened out from the 22 ESI subject ranking data of several domestic universities in the past two years as prediction objects.

[0023] Step 2: Divide the prediction object into training set and test set according to the set ratio, and customize the calculate function; In a preferred but non-limiting embodiment of the present invention, in step 2, the ratio is set to 4:1, that is, the training set is 4 times the test set.

[0024] like Figure 3 As shown, in a preferred but non-limiting embodiment of the present invention, step 2 specifically further includes: Convert the training set and test set as the target variable into classification format, that is, classify the target variable; The categories obtained by classifying the target variable are calculated to obtain their category weights to handle imbalanced data sets.

[0025] In a preferred but non-limiting embodiment of the present invention, in step 2, a custom calculate function is used to calculate the error, that is, to calculate the error between the predicted value and the true value.

[0026] like Figure 4 As shown, in a preferred but non-limiting embodiment of the present invention, in step 2, the custom calculate function includes: Calculate the mean absolute error MAE, which is the average absolute value of the difference between the predicted value and the actual value; Calculate the variance of the mean absolute error MAE to measure the volatility of the variance; Returns the calculated mean absolute error (MAE) and its variance.

[0027] Step 3: Construct a model for the development trend of ESI disciplines in universities; like Figure 5 As shown, in a preferred but non-limiting embodiment of the present invention, in step 3, the model for the development trend of ESI disciplines in colleges and universities is constructed, specifically including: Forecasting is performed using twelve different regression methods, namely: Perform forecasts using a linear regression model; Predictions were performed using a ridge regression model; Predictions were performed using the lasso regression model; Forecasting was performed using the elastic net regression model; Predictions were performed using a support vector regression model; Perform predictions using a decision tree regression model; Predictions were performed using a random forest regression model; Perform predictions using a gradient boosting regression model; The prediction is performed using the AdaBoost regression model; Use XGBoost regression model to perform prediction; Use CatBoost regression model to perform prediction; Use a multi-layer perceptron regression model to perform predictions; The model built for the development trend of ESI disciplines in colleges and universities also includes: Use the twelve regression methods one by one to perform prediction and store the prediction results in a dictionary; Returns the prediction results obtained by performing prediction using the twelve regression methods described above.

[0028] Step 4: Use the model for the development trend of ESI disciplines in colleges and universities to train the model; like Figure 6 As shown, in a preferred but non-limiting embodiment of the present invention, in step 4, using The method of model training for the model of ESI discipline development trend in colleges and universities specifically includes: Send the training set to the model for the development trend of ESI disciplines in colleges and universities to train the model and obtain a trained model; Use the trained model to predict the next data point to get the predicted value; Use the custom calculate function to obtain the mean absolute error MAE of the training set and its variance; Returns the unnormalized forecast value, the mean absolute error (MAE), and the variance of the MAE.

[0029] Step 5: Perform model prediction using the specified regression model.

[0030] like Figure 7 As shown, in a preferred but non-limiting embodiment of the present invention, in step 5, the scheme of executing model prediction by a specified regression model specifically includes: Get the newly prepared ESI subject ranking data; Use the trained model to predict the newly prepared ESI subject ranking data, and use the custom calculate function to obtain the mean absolute error MAE and the variance of the mean absolute error MAE of the newly prepared ESI subject ranking data; Returns the predicted value, mean absolute error (MAE) and the variance of the mean absolute error (MAE) of the newly prepared ESI subject ranking data.

[0031] like Figure 2 As shown, the information management device for intelligent archives described in the present invention includes: Processing terminal. The modules running on the processing terminal include: Model partitioning module, model building module, model training module and model prediction module; The model partitioning module is used to divide the prediction object into training set and test set according to the set ratio, and also customize the calculate function; The model building module is used to build a model for the development trend of ESI disciplines in colleges and universities; The model training module is used to train models using models for the development trends of ESI disciplines in universities; The model prediction module is used to perform model prediction through a specified regression model. The processing terminal can be a computer.

[0032] Specifically, the specific technical solution of the present invention mainly includes: 1. Collect 22 ESI subject ranking data from more than 100 domestic universities in the past two years.

[0033] 2. Select the global citation volume of 1% of 22 ESI disciplines, the global citation volume of 1% of 1,000% of 1,000% of 1, and the global citation volume of the university itself as the prediction objects.

[0034] 3. Divide the training set and test set into a ratio of 4:1.

[0035] 4. Build a fusion model (a model for the development trend of ESI disciplines in colleges and universities).

[0036] 5. Train the model.

[0037] 6. Such as Figures 8 to 10 As shown in the figure, the accuracy of the test model is above 90%.

[0038] The device of the present invention is composed of a model partitioning module, a model building module, a model training module and a model prediction module. The device constructs deep learning models of different architectures, performs large-scale data training, and is ultimately used for resource education subject classification tasks.

[0039] Key features: Custom calculate function: Define and implement a custom error evaluation function to improve the model's ability to handle unbalanced data.

[0040] Model construction: Twelve different model architectures are integrated to meet the needs of different tasks.

[0041] Model training: By defining training parameters (such as training cycle, batch size, early stopping mechanism, etc.), using Adam optimizer and class weight balancing technology, the model training process is optimized and the best model is saved.

[0042] Model prediction: load the trained model, make predictions on new data, and return the prediction results.

[0043] The beneficial effects of the present invention are that, compared with the prior art, the technical effects of the present invention include: It improves the shortcomings of domestic assessment of the development trend of ESI disciplines in colleges and universities, and helps colleges and universities calculate the time node when ESI disciplines enter the global 1%, as well as the possibility and time period of continuing to move towards the global 1 / 1000 and 1 / 10,000. Efficient model training: By customizing the error evaluation function and balancing the category weights, the performance of the model can be improved when processing unbalanced data. Multi-model architecture: It integrates twelve different model architectures and can adapt to complex data requirements. Automated model saving and loading: The system automatically saves the best model and automatically loads the model during prediction, simplifying the model management process.

[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not deviate from the spirit and scope of the present invention should be covered within the protection space of the claims of the present invention.

Claims

1. An evaluation and prediction method for the development trend of ESI disciplines in colleges and universities, characterized by: include: Step 1: Collect the 22 ESI subject ranking data of several domestic universities in the past two years to confirm the prediction object; Step 2: Divide the prediction object into training set and test set according to the set ratio, and customize the calculate function; Step 3: Construct a model for the development trend of ESI disciplines in universities; Step 4: Use the model for the development trend of ESI disciplines in colleges and universities to train the model; Step 5: Perform model prediction using the specified regression model.

2. The method for evaluating and predicting the development trend of ESI disciplines in colleges and universities according to claim 1 is characterized in that: In step 1, the number of colleges and universities is more than 100.

3. The method for evaluating and predicting the development trend of ESI disciplines in colleges and universities according to claim 2 is characterized in that: In step 1, the global citation volume of 1%, 1 / 1000%, 1 / 10,000 and the citation volume of the universities themselves of 22 ESI disciplines are selected from the ESI discipline ranking data of several domestic universities in the past two years as prediction objects.

4. The method for evaluating and predicting the development trend of ESI disciplines in colleges and universities according to claim 3 is characterized in that: In step 2, the ratio is set to 4:1, which means that the training set is 4 times the test set.

5. The method for evaluating and predicting the development trend of ESI disciplines in colleges and universities according to claim 4 is characterized in that: Step 2 specifically also includes: Convert the training set and test set as the target variable into classification format, that is, classify the target variable; The categories obtained by classifying the target variable are calculated to obtain their category weights to handle imbalanced data sets.

6. The method for evaluating and predicting the development trend of ESI disciplines in colleges and universities according to claim 5 is characterized in that: In step 2, the custom calculate function is used to calculate the error, that is, to calculate the error between the predicted value and the true value.

7. The method for evaluating and predicting the development trend of ESI disciplines in colleges and universities according to claim 6 is characterized in that: In step 2, the custom calculate function includes: Calculate the mean absolute error MAE, which is the average absolute value of the difference between the predicted value and the actual value; Calculate the variance of the mean absolute error MAE to measure the volatility of the variance; Returns the calculated mean absolute error (MAE) and its variance.

8. The method for evaluating and predicting the development trend of ESI disciplines in colleges and universities according to claim 7 is characterized in that: In step 3, the model for the development trend of ESI disciplines in colleges and universities is constructed, which specifically includes: Forecasting is performed using twelve regression methods, namely: Perform forecasts using a linear regression model; Predictions were performed using a ridge regression model; Predictions were performed using the lasso regression model; Forecasting was performed using the elastic net regression model; Predictions were performed using a support vector regression model; Perform predictions using a decision tree regression model; The prediction was performed using a random forest regression model; Perform predictions using a gradient boosting regression model; The prediction is performed using the AdaBoost regression model; Use XGBoost regression model to perform prediction; Use CatBoost regression model to perform prediction; Use a multi-layer perceptron regression model to perform predictions; The model built for the development trend of ESI disciplines in colleges and universities also includes: Use the twelve regression methods one by one to perform prediction and store the prediction results in a dictionary; Returns the prediction results obtained by performing prediction using the twelve regression methods described above.

9. The method for evaluating and predicting the development trend of ESI disciplines in colleges and universities according to claim 8 is characterized in that The feature is that in step 4, the method of using the model for the development trend of ESI disciplines in colleges and universities to train the model specifically includes: Send the training set to the model for the development trend of ESI disciplines in colleges and universities to train the model and obtain a trained model; Use the trained model to predict the next data point to get the predicted value; Use the custom calculate function to obtain the mean absolute error MAE of the training set and its variance; Returns the inverse normalized forecast value, mean absolute error (MAE), and the variance of the mean absolute error (MAE); In step 5, the model prediction scheme is executed through the specified regression model, including: Get the newly prepared ESI subject ranking data; Use the trained model to predict the newly prepared ESI subject ranking data, and use the custom calculate function to obtain the mean absolute error MAE and the variance of the mean absolute error MAE of the newly prepared ESI subject ranking data; Returns the predicted value, mean absolute error (MAE) and the variance of the mean absolute error (MAE) of the newly prepared ESI subject ranking data.

10. An information management device for intelligent archives, characterized in that: include: Processing terminal. The modules running on the processing terminal include: Model partitioning module, model building module, model training module and model prediction module; The model partitioning module is used to divide the prediction object into training set and test set according to the set ratio, and also customize the calculate function; The model building module is used to build a model for the development trend of ESI disciplines in colleges and universities; The model training module is used to train models using models for the development trends of ESI disciplines in universities; The model prediction module is used to perform model prediction using a specified regression model.

Images