Instrument sharing platform operation effect prediction method based on machine learning algorithm

Abstract

The invention provides an instrument sharing platform operation effect prediction method based on a machine learning algorithm. According to the technical scheme, the method comprises the following steps: obtaining operation data of an instrument sharing platform; screening operation data influence factors of the instrument sharing platform, and establishing an operation effect prediction model of the instrument sharing platform based on a machine learning algorithm; according to the established operation effect model of the instrument sharing platform, the operation effect of the instrument sharing platform is predicted and analyzed, and a final prediction result is obtained and serves as an operation effect result of the instrument sharing platform. The optimal machine learning algorithm is obtained through a large amount of calculation, the instrument sharing platform operation effect prediction method based on the machine learning algorithm is designed, the instrument sharing platform operation effect is predicted through the machine learning algorithm, and the instrument sharing platform operation effect can be accurately predicted.

Description

Technical Field

[0001] The present invention relates to the cross-technical field of artificial intelligence and instrument management, and more specifically, to a method for predicting the operating effectiveness of an instrument sharing platform based on a machine learning algorithm. Background Art

[0002] With the advancement of scientific capabilities and the increasing maturity of artificial intelligence (AI) technology, scientific and technological instruments have attracted considerable public interest. In addition, large instruments and equipment are basic resources for university teaching and research, and are also important indicators for measuring the academic and research capabilities of universities. With the continuous increase in investment in higher education, more and more large instruments and equipment are purchased by colleges and universities. At the same time, national and local governments have introduced diversified management models for open sharing of major scientific research infrastructure and large instruments. Many domestic colleges and universities and research institutions have also established a large number of instrument sharing platforms to improve resource utilization. The report shows that open equipment sharing has greatly increased the number of users and overall service time. However, challenges in equipment failure, laboratory safety risks and user appointment management have also increased significantly, which poses a huge obstacle to the efficient operation of instrument sharing platforms.

[0003] The rapid development of instrument sharing platforms requires intelligent management and operation modes. Researchers have explored effective solutions to address challenges in equipment failure frequency, laboratory safety risks, and user appointment management. The OGSA grid architecture has been used in major instrument networks to promote the sharing and collaborative use of large-scale equipment resources to maximize efficiency. Taking China as an example, the construction of a scientific instrument public service platform based on the O2O model is explained. The platform integrates scientific resources and research services, improves resource utilization, reduces equipment failure rate, and improves overall sharing efficiency. In order to improve the accuracy of scientific research instrument management and better support scientific research needs, an intelligent management platform for scientific research instruments based on the Internet of Things is proposed. The Java-based instrument sharing platform effectively optimizes instrument use, reduces laboratory safety risks, promotes sustainable development of science and technology, and expands resource sharing capabilities. However, these successful platform frameworks and management models have not yet been fully applied to the efficient operation of large instruments and equipment. Considering the important role of factors such as instrument management personnel, instrument type, pricing, test time, service level, and maintenance, it is crucial to balance instrument operation and management.

[0004] As a basic branch of artificial intelligence (AI), machine learning (ML) has shown great application potential in different fields such as physics, metamaterials, medicine, smart materials, education, computing and mathematics; in recent years, the rapid development of AI and data mining technology has promoted this application. For example, a ML-based model was developed to explore the relationship between the flame retardancy of organophosphorus flame retardants (OPFRs) and their structural characteristics, so that the feature importance can be analyzed more effectively. Combining ML technology with management research, especially in the management of large instruments and equipment, can derive operation rules from a large amount of instrument operation data. Data-driven mathematical models are a new paradigm for studying efficient management and operation strategies of large instruments and equipment. However, there is no report on the use of machine learning algorithms to predict the operating efficiency of instrument sharing platforms, which makes it impossible for existing technologies to quickly obtain instrument sharing platform operating efficiency data, reducing the accuracy of the obtained instrument sharing platform operating efficiency. Based on this, it is currently necessary to study a new prediction method to accurately predict the operating efficiency of instrument sharing platforms. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention proposes a method for predicting the operating effectiveness of an instrument sharing platform based on a machine learning algorithm, which realizes the prediction of the operating effectiveness of the instrument sharing platform using a machine learning algorithm. It can not only quickly obtain the operating effectiveness data of the instrument sharing platform and improve the accuracy of the obtained operating effectiveness data of the instrument sharing platform, but also provide important technical support for realizing the efficient operation and high-quality development of the instrument sharing platform.

[0006] A method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm is provided. The technical solution is as follows:

[0007] St10, obtain the operation data of the instrument sharing platform;

[0008] St20. Screen the influencing factors of the operation data of the instrument sharing platform and establish a prediction model for the operation effect of the instrument sharing platform based on the machine learning algorithm;

[0009] St30. According to the established instrument sharing platform operation effectiveness model, the instrument sharing platform operation effectiveness is predicted and analyzed, and the final prediction result is obtained as the instrument sharing platform operation effectiveness result.

[0010] To sum up, the above technical scheme has the following beneficial effects: the present invention obtains the optimal machine learning algorithm through a large number of calculations, designs a method for predicting the operating effectiveness of an instrument sharing platform based on a machine learning algorithm, and realizes the use of a machine learning algorithm to predict the operating effectiveness of the instrument sharing platform. The operating performance of the instrument sharing platform can be accurately predicted, and finally the operating performance parameters of the instrument sharing platform can be obtained. It can not only quickly obtain the operating performance data of the instrument sharing platform and improve the accuracy of the obtained operating performance data of the instrument sharing platform, but also provide important technical support for realizing the efficient operation and high-quality development of the instrument sharing platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 Flowchart for using machine learning algorithms to predict test costs from running an instrument sharing platform.

[0012] Figure 2 The actual value-predicted value graph and residual graph of the test fee obtained by the prediction analysis of the support vector regression SVR instrument sharing platform operation effectiveness prediction model in Example 1.

[0013] Figure 3 It is the actual value-predicted value graph and residual graph of the test fee obtained by the prediction analysis of the artificial neural network ANN instrument sharing platform operation effectiveness prediction model in Example 2.

[0014] Figure 4 The actual value-predicted value graph and residual graph of the test fee obtained by the prediction analysis of the Gaussian process regression GPR instrument sharing platform operation effectiveness prediction model in Example 3.

[0015] Figure 5 It is the actual value-predicted value graph and residual graph of the test fee obtained by the prediction analysis of the decision tree DT instrument sharing platform operation effectiveness prediction model in Example 4.

[0016] Figure 6 The actual value-predicted value graph and residual graph of the test fee obtained by the prediction analysis of the random forest RF instrument sharing platform running effectiveness prediction model in Example 5. DETAILED DESCRIPTION

[0017] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] A method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm includes the following process:

[0019] St10, obtain the operation data of the instrument sharing platform;

[0020] St20. Screen the influencing factors of the operation data of the instrument sharing platform and establish a prediction model for the operation effect of the instrument sharing platform based on the machine learning algorithm;

[0021] St30. According to the established instrument sharing platform operation effectiveness model, the instrument sharing platform operation effectiveness is predicted and analyzed, and the final prediction result is obtained as the instrument sharing platform operation effectiveness result.

[0022] The present invention obtains the optimal machine learning algorithm through a large number of calculations, designs a method for predicting the operating effectiveness of an instrument sharing platform based on the machine learning algorithm, and realizes the use of the machine learning algorithm to predict the operating effectiveness of the instrument sharing platform. The operating effectiveness of the instrument sharing platform can be accurately predicted, and finally the operating effectiveness parameters of the instrument sharing platform can be obtained. It can not only quickly obtain the operating effectiveness data of the instrument sharing platform and improve the accuracy of the obtained operating effectiveness data of the instrument sharing platform, but also provide important technical support for realizing the efficient operation and high-quality development of the instrument sharing platform.

[0023] St11. Filter the instrument operation data of the instrument sharing platform in a certain period of time and export it to obtain the instrument sharing platform operation data. The instrument sharing platform operation data is obtained by intercepting a certain period of time. By intercepting the operation data of different time periods, multiple original data can be obtained, which is conducive to establishing an instrument sharing platform operation effect prediction model.

[0024] The factors affecting the operation data of the instrument sharing platform include operation time, instrument type, unit price and test time.

[0025] The establishment of an instrument sharing platform operation performance prediction model based on machine learning algorithms includes the following processes:

[0026] St21, using operating time, instrument type, unit price and test time as input parameters and test fee as output parameter, the data are imported into the software and calculated through machine learning algorithm to establish a prediction model for the operation effect of the instrument sharing platform;

[0027] St22, through the determination coefficient R of the prediction model 2 , mean square error (MSE), root mean square error (RMSE) and mean absolute error (MAE) are used to evaluate the performance of the instrument sharing platform operation effectiveness prediction model.

[0028] The machine learning algorithms include artificial neural network ANN, Gaussian process regression GPR, decision tree DT, random forest RF and support vector regression SVR. The performance of the instrument sharing platform operation effect prediction model obtained by different algorithms is different. Therefore, it is necessary to evaluate the instrument sharing platform operation effect prediction model obtained by different algorithms to obtain the best algorithm model. Preferably, the machine learning algorithm is a support vector regression model.

[0029] The performance evaluation index of the prediction model is the determination coefficient R 2 Calculated by the following process:

[0030]

[0031] In the above error calculation: t i is the actual output value, td i is the predicted value of the machine learning algorithm model, N is the number of data, is the average value.

[0032] The mean square error (MSE) is calculated by the following process:

[0033]

[0034] In the above error calculation: n is the number of samples, y is the true value, and y′ is the predicted value.

[0035] The root mean square error RMSE is calculated by the following process:

[0036]

[0037] In the above error calculation: N is the number of data, t i is the actual output value, td i Predict values ​​for machine learning algorithm models;

[0038] The mean absolute error (MAE) is calculated by the following process:

[0039]

[0040] In the above error calculation: n is the number of samples, y is the true value, and y′ is the predicted value.

[0041] Example 1: Use support vector regression SVR algorithm to establish an instrument sharing platform operation effectiveness prediction model based on machine learning algorithm. A database is built with 3710 instrument operation data from January 2023 to August 2024 of an instrument sharing platform in a certain university in Zhejiang Province. The influencing factors such as operation time (month), instrument type, unit price (yuan / hour) and test time (hour) are used as input parameters, and the support vector regression (SVR) model is established with test fee (yuan) as the output variable. The selected influencing factor data and test fee data are integrated into an operation effectiveness data set, which is randomly divided into a training set and a test set in a ratio of 4:1 after normalization, and the machine learning model operation is completed by genetic algorithm and 5-fold cross validation algorithm. The material data in the training set is used for model training, and the material data in the test set is used to test the accuracy and generalization ability of the model.

[0042] The specific data obtained by analyzing the support vector regression (SVR) algorithm instrument sharing platform operation effect prediction model are shown in Table 1. The actual value-predicted value graph and residual graph of the instrument sharing platform test fee are shown in Table 1. Figure 2 shown.

[0043] Example 2: The difference from Example 1 is that this example adopts the second machine learning algorithm for verification, and uses the artificial neural network ANN algorithm to establish an instrument sharing platform operation effectiveness prediction model. Taking the influencing factors such as operating time (month), instrument type, unit price (yuan / hour) and test time (hours) as input parameters, and the test fee (yuan) as the output variable to establish an artificial neural network ANN model, the data preprocessing, data set division method, and model evaluation index are consistent with Example 1. The specific data obtained by analyzing the artificial neural network ANN instrument sharing platform operation effectiveness prediction model are shown in Table 1, and the instrument sharing platform test fee actual value-predicted value graph and residual graph are shown in Table 1. Figure 3 shown.

[0044] Example 3: The difference from Example 1 is that this example adopts the third machine learning algorithm for verification, and uses the decision tree DT algorithm to establish a prediction model for the operation effectiveness of the instrument sharing platform. Taking the influencing factors such as operation time (month), instrument type, unit price (yuan / hour) and test time (hours) as input parameters, and the test fee (yuan) as the output variable to establish a decision tree GPR model, the data preprocessing, data set division method, and model evaluation index are consistent with Example 1. The specific data obtained by analyzing the Gaussian regression process GPR instrument sharing platform operation effectiveness prediction model are shown in Table 1, and the instrument sharing platform test fee true value-predicted value graph and residual graph are shown in Table 1. Figure 4 shown.

[0045] Example 4: The difference from Example 1 is that this example adopts the fourth machine learning algorithm for verification, and uses the decision tree DT algorithm to establish a prediction model for the operation effectiveness of the instrument sharing platform: the operating time (month), instrument type, unit price (yuan / hour) and test time (hour) and other influencing factors are used as input parameters, and the test fee (yuan) is used as the output variable to establish a decision tree DT model. The data preprocessing, data set division method, and model evaluation index are consistent with Example 1. The specific data obtained by analyzing the decision tree DT instrument sharing platform operation effectiveness prediction model are shown in Table 1, and the instrument sharing platform test fee true value-predicted value graph and residual graph are shown in Table 1. Figure 5 shown.

[0046] Example 5: The difference from Example 1 is that this example adopts the fifth machine learning algorithm for verification, and uses the random forest RF algorithm to establish a prediction model for the operation effectiveness of the instrument sharing platform. Taking the influencing factors such as operation time (month), instrument type, unit price (yuan / hour) and test time (hours) as input parameters, and the test fee (yuan) as the output variable to establish a random forest RF model, the data preprocessing, data set division method, and model evaluation indicators are consistent with Example 1. The specific data obtained by analyzing the random forest RF instrument sharing platform operation effectiveness prediction model are shown in Table 1, and the actual value-predicted value graph and residual graph of the instrument sharing platform test fee are shown in Table 1. Figure 6 shown.

[0047] The determination coefficient R of the prediction model 2 , mean square error (MSE), root mean square error (RMSE) and mean absolute error (MAE) are used to evaluate the performance of the instrument sharing platform operation effectiveness prediction model obtained by the five machine learning algorithms. Table 1 shows the statistical results of the five machine learning algorithms for predicting the operation effectiveness of the instrument sharing platform.

[0048] Table 1 Prediction results of machine learning model for instrument sharing platform operation performance

[0049] SVR ANN GPR DT RF <![CDATA[R 2 ]]> 0.98 0.97 0.97 0.87 0.81 MSE 1.94E+05 2.68E+05 2.79E+05 1.08E+06 1.62E+06 RMSE 440.4 517.63 528.1 1039.6 1274.5 MAE 226.15 257.77 214.38 458.36 510.94

[0050] In summary, this application lists five machine learning algorithms in the prediction methods of instrument sharing platform operation effectiveness (test fee), all of which show good prediction results. Compared with artificial neural network ANN, Gaussian process regression GPR, decision tree DT and random forest RF, support vector regression SVR has a higher prediction effect. Therefore, the optimal model algorithm is selected as support vector regression SVR.

[0051] The machine learning algorithm is the support vector regression model, which is established and operated through the following process:

[0052] St23, import the running time, instrument type, unit price, test time and test fee into csv file and then import into the advanced data structure;

[0053] St24, based on the operation time, instrument type, unit price, test time and test fee data, mapping and fusion are performed to form initial test fee analysis data;

[0054] St25, the initial test fee analysis data processed by St24 is divided into a test fee training data set and a test fee test data set according to a ratio of 4:1, and the cross-validation method is used to eliminate bad pixels;

[0055] St26, normalize the data after removing bad pixels;

[0056] St27, using an improved genetic algorithm to optimize the parameters of the support vector regression model, and obtain the optimal parameter penalty factor C and kernel function width coefficient σ;

[0057] St28, applying the penalty factor C and kernel function width coefficient σ obtained above to the support vector regression algorithm to predict the data to be tested;

[0058] Specifically, the method of normalization in St26 is: normalize the data to between [0,1],

[0059]

[0060] Among them, X normalize is the normalized data, X is the unnormalized data, and X min is the minimum value of the input data, X man is the maximum value of the input data.

[0061] Specifically, the parameter optimization method in process St27 is to use genetic algorithm and 5-fold cross validation algorithm to optimize parameters. The genetic algorithm uses the average of the square sum of the 5-fold cross validation errors as the fitness value and the regression coefficient R 2 To find the optimal parameters for the goal, the genetic algorithm parameters are population size of 200, chromosome number of 2, chromosome length of 20, maximum parameter value of 15, minimum parameter value of 0.001, and number of iterations of 200.

[0062] 7. According to claim 6, a method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm is characterized in that predicting the data to be measured in St28 includes the following process:

[0063] St281, using the support vector regression model for simulation prediction, in order to make regression predictions for liquidity, by constructing and solving the problem for optimization:

[0064]

[0065] Where R is the multiple correlation coefficient, ξ i and is a slack variable, w and b are model parameters, C is a parameter penalty factor, l is the number of data samples, 0≤i≤l, ε is an insensitive influence factor, x i is the feature of each sample, y i is the label of each sample;

[0066] St282. Construct and solve the dual problem of the original optimization problem:

[0067]

[0068] Get the optimal solution: where α and α * is the solution to the dual problem, l is the number of samples, K(x i ·x j ) is the kernel function, l is the number of data samples, 0≤j≤l, ε is the insensitive influence factor, y i is the label of each sample;

[0069] St283, select RBF kernel function:

[0070]

[0071] Among them, σ is the kernel function width coefficient, x i is the value of the input space, x j To distinguish from x i Another value in the input space. RBF kernel function, also known as radial basis function kernel function, is one of the most commonly used kernel functions in support vector machines.

[0072] St284, find the decision function and parameter b,

[0073]

[0074] The data to be tested x includes the operating time, instrument type, unit price and test time. Substituting the data to be tested x into the above formula, the test cost value f(x) can be obtained.

[0075] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm, characterized in that: The process includes: St10, obtain the operation data of the instrument sharing platform; St20. Screen the influencing factors of the operation data of the instrument sharing platform and establish a prediction model for the operation effect of the instrument sharing platform based on the machine learning algorithm; St30. According to the established instrument sharing platform operation effectiveness model, the instrument sharing platform operation effectiveness is predicted and analyzed, and the final prediction result is obtained as the instrument sharing platform operation effectiveness result.

2. According to claim 1, a method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm is characterized in that: The process includes: St11. Filter the instrument operation data of the instrument sharing platform for a certain period of time, and export it to obtain the instrument sharing platform operation data.

3. The method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm according to claim 1, characterized in that: The process includes: The operating data influencing factors of the instrument sharing platform include operating time, instrument type, unit price and test time.

4. The method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm according to claim 3 is characterized in that: The establishment of an instrument sharing platform operation performance prediction model based on machine learning algorithms includes the following processes: St21, using operating time, instrument type, unit price and test time as input parameters and test fee as output parameter, the data are imported into the software and calculated through machine learning algorithm to establish a prediction model for the operation effect of the instrument sharing platform; St22, through the determination coefficient R of the prediction model 2 , mean square error (MSE), root mean square error (RMSE) and mean absolute error (MAE) are used to evaluate the performance of the instrument sharing platform operation effectiveness prediction model.

5. The method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm according to claim 4, characterized in that: The performance evaluation index of the prediction model is the determination coefficient R 2 Calculated by the following process: In the above error calculation: t i is the actual output value, td i is the predicted value of the machine learning algorithm model, N is the number of data, is the average value.

6. The method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm according to claim 4, characterized in that: The mean square error MSE is calculated by the following process: In the above error calculation: n is the number of samples, y is the true value, and y′ is the predicted value.

7. The method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm according to claim 4, characterized in that: The root mean square error RMSE is calculated by the following process: In the above error calculation: N is the number of data, t i is the actual output value, td i Predict values ​​for machine learning algorithm models.

8. The method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm according to claim 4, characterized in that: The mean absolute error MAE is calculated by the following process: In the above error calculation: n is the number of samples, y is the true value, and y′ is the predicted value.

9. The method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm according to claim 3, characterized in that: The machine learning algorithm is a support vector regression model, which is established and operated through the following process: St23, import the running time, instrument type, unit price, test time and test fee into csv file and then import into the advanced data structure; St24, based on the operation time, instrument type, unit price, test time and test fee data, mapping and fusion are performed to form initial test fee analysis data; St25, the initial test fee analysis data processed by St24 is divided into a test fee training data set and a test fee test data set according to a ratio of 4:1, and the cross-validation method is used to eliminate bad pixels; St26, normalize the data after removing bad pixels; St27, using an improved genetic algorithm to optimize the parameters of the support vector regression model, and obtain the optimal parameter penalty factor C and kernel function width coefficient σ; St28. Apply the penalty factor C and kernel function width coefficient σ obtained above to the support vector regression algorithm to predict the data to be tested.

10. The method for predicting the operation effectiveness of an instrument sharing platform based on a machine learning algorithm according to claim 9, characterized in that: The prediction of the test data in St28 includes the following process: St281, using the support vector regression model for simulation prediction, in order to make regression predictions for liquidity, by constructing and solving the problem for optimization: Where R is the multiple correlation coefficient, ξ i and is a slack variable, w and b are model parameters, C is a parameter penalty factor, l is the number of data samples, 0≤i≤l, ε is an insensitive influence factor, x i is the feature of each sample, y i is the label of each sample; St282. Construct and solve the dual problem of the original optimization problem: Get the optimal solution: where α and α * is the solution to the dual problem, l is the number of data samples, K(x i ·x j ) is the kernel function, l is the number of data samples, 0≤j≤l, ε is the insensitive influence factor, y i is the label of each sample; St283, select RBF kernel function: Among them, σ is the kernel function width coefficient, x i is the value of the input space, x j To distinguish from x i Another value for the input space; St284, find the decision function and parameter b, The data to be tested x includes the operating time, instrument type, unit price and test time. Substituting the data to be tested x into the above formula, the test cost value f(x) can be obtained.

Images