A machine learning-based asphalt oil source identification model and automated implementation method

Abstract

A machine learning-based asphalt oil source identification model and automated implementation method, belonging to the field of automated asphalt oil source identification. The method comprises the following steps: constructing an asphalt data set based on the infrared spectral features and oil sources of different asphalt samples; proposing a data augmentation method suitable for the asphalt data set; extracting features of the infrared spectral features of asphalt based on the idea of ​​feature dimensionality reduction to determine the final features of the model; constructing an asphalt oil source identification model based on support vector machines, artificial neural networks and decision trees; determining the best asphalt oil source identification model; and designing a visual operation platform for the asphalt oil source identification model. The method realizes the accurate establishment of the asphalt oil source identification model and the automatic prediction of the asphalt oil source, effectively accelerating the speed and accuracy of asphalt oil source identification.

Description

Technical Field

[0001] The present invention belongs to the field of automated identification of asphalt oil sources, and in particular relates to an asphalt oil source identification model based on machine learning and an automated implementation method. Background Art

[0002] Asphalt is a common road material, and its performance is crucial to road quality. Asphalt performance is highly correlated with the oil source. Therefore, how to identify the asphalt oil source to select high-quality asphalt materials has become a key link in actual engineering construction. Traditional methods for identifying asphalt oil sources include density measurement, viscosity measurement, solubility measurement, and elemental analysis. However, the chemical composition of asphalt is complex, and these methods have some defects, such as high sample processing requirements, cumbersome operations, and long time consumption, making it difficult to quickly and accurately identify asphalt materials.

[0003] In order to overcome these problems, infrared spectroscopy came into being. This method provides information about the molecular structure and chemical composition of asphalt by measuring the absorption spectrum of asphalt in the infrared light region. Compared with traditional methods, it has become a powerful tool for studying the chemical composition and performance characteristics of asphalt due to its advantages such as non-destructive, rapid, diverse, highly sensitive and accurate. It also provides an effective means for quality control and material research and development in asphalt engineering and related industries. However, the existing asphalt oil source identification models are all based on traditional statistical methods, which have problems such as low accuracy, poor adaptability and cumbersome calculation.

[0004] In recent years, with the strong rise of artificial intelligence, analytical detection methods based on the coupling of artificial intelligence and infrared spectroscopy technology have been used in many scientific fields. Machine learning is a branch of artificial intelligence. Its goal is to enable computers to recognize patterns, make predictions and decisions from data by building and training models without explicit programming instructions. Taking infrared spectral characteristic parameters as model input parameters and asphalt oil source as output parameters, and making full use of the advantages of machine learning automation, high precision and high adaptability, the establishment of an asphalt oil source identification model based on machine learning is expected to achieve the goal of rapid identification of asphalt oil sources. Summary of the invention

[0005] The purpose of the present invention is to provide an asphalt oil source identification model based on machine learning and an automated implementation method to address the problems of the existing asphalt oil source identification methods, such as cumbersome operation, long time consumption, low identification accuracy, and high sample processing requirements. The method establishes a machine learning identification model for asphalt oil sources and builds a corresponding visual operation interface, achieving the goal of automatically predicting asphalt oil sources by inputting asphalt infrared spectrum parameters, effectively accelerating the speed of asphalt quality identification and improving the identification accuracy.

[0006] The objective of the present invention is achieved through the following technical solutions:

[0007] A machine learning-based asphalt oil source identification model and automated implementation method, the method comprising the following steps:

[0008] Step 1: Collect asphalt samples from different oil sources, test the infrared spectrum of asphalt using an infrared spectrometer, select the characteristic peak height and peak area of ​​the infrared spectrum as features, and preliminarily construct an asphalt dataset with a one-to-one correspondence between infrared spectrum features and oil sources;

[0009] Step 2: Select the SMOTE algorithm as the data augmentation method for expanding the asphalt dataset. Based on the data augmentation method, expand the number of samples in the original asphalt dataset to optimize the dataset structure and obtain the final asphalt dataset, thereby improving the generalization ability and performance of the model; the data augmentation method is the SMOTE algorithm because of its good balancing effect, low risk of avoiding overfitting and moderate computational complexity. The principle of the SMOTE algorithm is to generate new synthetic samples between minority class samples to balance the dataset in order to achieve the purpose of data augmentation.

[0010] Step 3: Based on the idea of ​​feature dimensionality reduction, a feature extraction method suitable for asphalt infrared spectrum is studied. The feature extraction method is used to process the asphalt infrared spectrum features to reduce the number of features, determine the final input parameters of the asphalt identification model, and ensure the performance of the identification model;

[0011] Step 4: Using the infrared spectral features determined in step 3 as the model input parameters and the asphalt oil source as the model output parameters, asphalt oil source identification models are constructed based on support vector machines, artificial neural networks, and decision trees, respectively. The above three algorithms are optimized for hyperparameters based on the grid search method and cross-validation method to determine the best hyperparameter combination for each model; hyperparameters are parameters that affect the model training process and final performance, and need to be manually set before training. For example, in the support vector machine (SVM), the regularization parameter and the kernel function parameter are hyperparameters. Different hyperparameter combinations will significantly affect the performance of the SVM model. By optimizing the hyperparameters, the optimal parameter combination can be found, thereby improving the prediction accuracy and generalization ability of the model.

[0012] Step 5: Based on the accuracy, precision, recall, F1 value and Kappa coefficient, the actual effects of the three asphalt oil source identification models constructed in step 4 are evaluated respectively, and the best asphalt oil source identification model is comprehensively determined;

[0013] Step 6: Build a visual operation platform for the asphalt oil source identification model based on the PyQT5 tool, and integrate the optimal feature extraction method of the asphalt data set determined in step 3 and the optimal asphalt oil source identification model determined in step 5 into the platform to realize the automatic identification of asphalt oil source by inputting asphalt infrared spectral characteristic parameters.

[0014] Furthermore, the feature extraction method suitable for asphalt infrared spectrum based on the feature dimension reduction idea is specifically as follows:

[0015] Step 31: Extract the features of infrared spectrum of asphalt based on principal component analysis;

[0016] Step 32: Extract the infrared spectrum characteristics of asphalt based on linear discriminant analysis;

[0017] Step 33: Visualize the feature extraction results of principal component analysis and linear discriminant analysis to determine the best feature extraction method for asphalt infrared spectral characteristics.

[0018] Furthermore, in the step 31, the principle of the principal component analysis method is to map high-dimensional data to low-dimensional space while retaining the main features and structure of the data as much as possible; in the step 32, the principle of the linear discriminant analysis method is to maximize the separability between different categories by finding linear combinations in the feature space; in the step 33, the visualization comparison method is to apply the KMeans clustering algorithm to the data after dimensionality reduction by principal component analysis and linear discriminant analysis, and set the same number of clusters to compare the clustering effects of the two dimensionality reduction methods under the same conditions.

[0019] Furthermore, in step 4, the basic idea of ​​the support vector machine is to find an optimal hyperplane so that the distance from the data points on both sides of the hyperplane to the hyperplane is maximized; the artificial neural network is a network structure composed of multiple neuron layers, which generates output by receiving input signals and performing weighted summation, and processing them with an activation function; the basic idea of ​​the decision tree is to make decisions by recursively dividing the feature space and constructing a tree; the main idea of ​​the grid search method and cross-validation method is to exhaustively search all possible parameter combinations and evaluate the performance of each parameter combination through cross-validation, which can systematically explore the parameter space, automatically find the best parameter combination, and reduce the workload and complexity of manual parameter adjustment.

[0020] Furthermore, in step five, the accuracy rate refers to the ratio of the number of samples correctly predicted by the model to the total number of samples; precision rate refers to the ratio of all samples predicted to be positive that are actually positive. The higher the precision rate, the fewer errors the model makes when predicting positive classes; recall rate refers to the ratio of all samples actually positive that are correctly predicted to be positive. The higher the recall rate, the more positive samples the model can identify; F1 value: the F1 value is the harmonic mean of the precision rate and the recall rate. When both indicators are high, the F1 value will also be high; Kappa coefficient: used to measure the consistency between the classification results of the classifier and the random classification results. The value range of the Kappa coefficient is [-1,1]. The closer it is to 1, the more consistent the classification results of the model are with the actual labels.

[0021] Furthermore, in the step six, first input all the asphalt infrared spectrum characteristic parameters; secondly, click the confirmation button to call the optimal feature extraction method of the asphalt data set determined in step three to automatically extract the infrared spectrum features; finally, call the optimal asphalt oil source identification model determined in step five and automatically output the asphalt oil source according to the features after feature extraction, so as to realize the automatic identification of the asphalt oil source.

[0022] Compared with the prior art, the present invention has the following advantages: the present invention proposes a data augmentation method suitable for asphalt data sets, constructs a reliable oil source identification model based on asphalt infrared spectrum characteristics, and builds a visual operation platform for the asphalt oil source identification model, achieving the goal of inputting asphalt infrared spectrum parameters and automatically and conveniently outputting asphalt oil sources. It effectively accelerates the speed and accuracy of asphalt oil source identification and promotes the development of the asphalt material quality identification industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flow chart of the asphalt oil source identification model and automated implementation method of the present invention;

[0024] Figure 2 This is the result of data dimensionality reduction based on principal component analysis;

[0025] Figure 3 This is the result diagram of data dimensionality reduction based on linear discriminant analysis;

[0026] Figure 4 This is a comparison chart of the index scores of the three asphalt oil source identification models of the present invention;

[0027] Figure 5 This is a diagram of the visual operating platform of the asphalt oil source identification model of the present invention. DETAILED DESCRIPTION

[0028] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention that does not depart from the spirit and scope of the technical solution of the present invention should be included in the protection scope of the present invention.

[0029] The present invention provides an asphalt oil source identification model based on machine learning and an automated implementation method. The method constructs an asphalt data set based on the infrared spectral features and oil sources of different asphalt samples; proposes a data augmentation method suitable for the asphalt data set; extracts the features of the asphalt infrared spectral features based on the idea of ​​feature dimensionality reduction, and determines the final features of the model; constructs an asphalt oil source identification model based on support vector machines, artificial neural networks and decision trees; determines the optimal asphalt oil source identification model; designs a visual operation platform for the asphalt oil source identification model, and realizes the automation of the model. Figure 1The method specifically comprises the following steps:

[0030] Step 1: Collect asphalt samples from different oil sources, test the infrared spectrum of asphalt based on infrared spectrometer, select the characteristic peak height and peak area of ​​infrared spectrum as features, and preliminarily construct an asphalt data set with one-to-one correspondence between infrared spectrum features and oil sources.

[0031] Step 2: Explore data augmentation methods suitable for asphalt datasets, expand the number of asphalt samples in the original dataset based on data augmentation methods to optimize the dataset structure, and obtain the final asphalt dataset, thereby improving the model's generalization ability and performance.

[0032] In this step, the data augmentation method is the SMOTE algorithm, because of its good balancing effect, low risk of avoiding overfitting and moderate computational complexity. The principle of the SMOTE algorithm is to generate new synthetic samples between minority class samples to balance the data set to achieve the purpose of data augmentation.

[0033] Step 3: Based on the idea of ​​feature dimensionality reduction, a feature extraction method suitable for asphalt infrared spectrum is studied. The feature extraction method is used to process the asphalt infrared spectrum features to reduce the number of features, determine the final input parameters of the asphalt identification model, and ensure the performance of the identification model:

[0034] Step 31: Extract the features of infrared spectrum of asphalt based on principal component analysis;

[0035] Step 32: Extract the infrared spectrum characteristics of asphalt based on linear discriminant analysis;

[0036] Step 33: Visualize the feature extraction results of principal component analysis and linear discriminant analysis to determine the best feature extraction method for asphalt infrared spectral characteristics.

[0037] In this step three, the principle of the principal component analysis method is to map high-dimensional data to a low-dimensional space while retaining the main features and structure of the data as much as possible.

[0038] In step 32, the principle of the linear discriminant analysis method is to maximize the separability between different categories by finding linear combinations in the feature space.

[0039] In this step three-three, the visual comparison method is to apply the KMeans clustering algorithm to the data after the principal component analysis and linear discriminant analysis dimensionality reduction, and set the same number of clusters to compare the clustering effects of the two dimensionality reduction methods under the same conditions.

[0040] Step 4: Using the infrared spectral features determined in step 3 as the model input parameters and the asphalt oil source as the model output parameters, asphalt oil source identification models are constructed based on support vector machine, artificial neural network and decision tree respectively, and the hyper-parameters of the above three algorithms are optimized based on grid search method and cross-validation method to determine the best hyper-parameter combination for each model.

[0041] In this step, the basic idea of ​​the support vector machine is to find an optimal hyperplane so that the distance from the data points on both sides of the hyperplane to the hyperplane is maximized; the artificial neural network is a network structure composed of multiple neuron layers, which generates output by receiving input signals and performing weighted summation, and processing them with an activation function; the basic idea of ​​the decision tree is to make decisions by recursively dividing the feature space and constructing a tree; the main idea of ​​the grid search method and cross-validation method is to exhaustively search all possible parameter combinations and evaluate the performance of each parameter combination through cross-validation, which can systematically explore the parameter space, automatically find the best parameter combination, and reduce the workload and complexity of manual parameter adjustment.

[0042] Step 5: Based on the accuracy, precision, recall, F1 value and Kappa coefficient, the actual effects of the three asphalt oil source identification models constructed in step 4 are evaluated respectively, and the best asphalt oil source identification model is comprehensively determined.

[0043] In this step, the accuracy rate refers to the ratio of the number of samples correctly predicted by the model to the total number of samples; precision rate refers to the ratio of all samples predicted to be positive that are actually positive. The higher the precision rate, the fewer errors the model makes when predicting positive classes; recall rate refers to the ratio of all samples actually positive that are correctly predicted to be positive. The higher the recall rate, the more positive samples the model can identify; F1 value: the F1 value is the harmonic mean of the precision rate and the recall rate. When both indicators are high, the F1 value will also be high; Kappa coefficient: used to measure the consistency between the classification results of the classifier and the random classification results. The value range of the Kappa coefficient is [-1,1]. The closer it is to 1, the more consistent the classification results of the model are with the actual labels.

[0044] Step 6: Build a visual operation platform for the asphalt oil source identification model based on the PyQT5 tool, and integrate the optimal feature extraction method of the asphalt data set determined in step 3 and the optimal asphalt oil source identification model determined in step 5 into the platform to realize the automatic identification of asphalt oil source by inputting asphalt infrared spectral characteristic parameters.

[0045] In this step, first input all the characteristic parameters of asphalt infrared spectrum; secondly, click the confirmation button to call the optimal feature extraction method of the asphalt data set determined in step three to automatically extract the infrared spectrum features; finally, call the optimal asphalt oil source identification model determined in step five and automatically output the asphalt oil source according to the features after feature extraction to realize the automatic identification of asphalt oil source.

[0046] Embodiment 1:

[0047] A machine learning-based asphalt oil source identification model and automated implementation method, the specific operation process is as follows:

[0048] Step 1: Collect a total of 56 asphalt samples from 7 different oil sources, test the infrared spectra of asphalt based on infrared spectrometer, select 22 typical infrared spectral characteristic peak heights and peak areas as features, and preliminarily construct an asphalt data set with infrared spectral features and oil sources corresponding one to one, as shown in Table 1.

[0049] Table 1 Asphalt dataset

[0050] Characteristic parameters L-1 L-2 … AM-1 AM-2 <![CDATA[I 2920 ]]> 0.179 0.139 … 0.145 0.172 <![CDATA[I 2850 ]]> 0.11 0.078 … 0.084 0.107 <![CDATA[I 1700 ]]> 0.006 0.005 … 0.004 0 <![CDATA[I 1600 ]]> 0.014 0.011 … 0.016 0.019 <![CDATA[I 1456 ]]> 0.086 0.083 … 0.082 0.095 <![CDATA[I 1375 ]]> 0.033 0.031 … 0.036 0.031 <![CDATA[I 1030 ]]> 0.004 0.008 … 0.006 0.019 <![CDATA[I 870 ]]> 0.01 0.01 … 0.008 0.018 <![CDATA[I 810 ]]> 0.008 0.016 … 0.012 0.036 <![CDATA[I 745 ]]> 0.004 0.007 … 0.01 0.011 <![CDATA[I 720 ]]> 0.012 0.01 … 0.01 0.018 <![CDATA[A 2920 ]]> 7.549 6.463 … 6.508 7.332 <![CDATA[A 2850 ]]> 2.513 2.034 … 2.055 2.393 <![CDATA[A 1700 ]]> 0.227 0 … 0.128 0 <![CDATA[A 1600 ]]> 0.834 0.108 … 0.614 0.618 <![CDATA[A 1456 ]]> 3.147 3.127 … 3.271 3.793 <![CDATA[A 1376 ]]> 0.939 0.819 … 0.753 0.793 <![CDATA[A 1030 ]]> 0.119 0.218 … 0.105 0.871 <![CDATA[A 870 ]]> 0.348 0.281 … 0.248 0.63 <![CDATA[A 810 ]]> 0.213 0.45 … 0.325 1.071 <![CDATA[A 745 ]]> 0.11 0.178 … 0.237 0.239 <![CDATA[A 720 ]]> 0.168 0.129 … 0.122 0.256

[0051] Note: In the table, I represents the absorption peak height, A represents the absorption peak area, and L-1, L-2, AM-1, AM-2, etc. represent the asphalt numbers.

[0052] Step 2: Based on the SMOTE algorithm, the number of asphalt samples in the original data set is expanded to optimize the data set structure and obtain the final asphalt data set. The data processing and model construction of the present invention are implemented on the JupyterNotebook platform based on the Python computer language. After data augmentation, the existing data is increased from 56 to 112, and the problem of sample imbalance is effectively solved. The data of the seven oil sources are 16 each. The comparison of the database structure before and after data augmentation is shown in Table 2 below.

[0053] Table 2 Comparison table before and after data augmentation

[0054]

[0055]

[0056] Step 3: Reduce the dimensions of the above 22 features based on principal component analysis and linear discriminant analysis, respectively. Apply the KMeans clustering algorithm to the data after principal component analysis and linear discriminant analysis, and set the same number of clusters to compare their clustering effects under the same conditions. By observing the visualization of the clustering results, you can intuitively see which dimensionality reduction method is more effective in retaining data structure and category information. Set the number of clusters to 7, and the results are as follows: Figure 2 and 3 shown.

[0057] from Figure 2 and 3 It can be seen that the points of different categories reduced by principal component analysis may overlap or cluster together and are not completely separated. In comparison, linear discriminant analysis performs better in category separation, and the boundaries between different categories are more obvious, so linear discriminant analysis is selected as the final feature extraction method. And when the linear discriminant analysis dimension reaches 4, the broken line reaches the highest point, which means that the model performance is optimal when the selected dimension is 4, so the optimal dimension is selected as 4, that is, the 22 feature parameters are reduced to 4 new parameters.

[0058] Step 4: Using the infrared spectral features determined in step 3 as the model input parameters and the asphalt oil source as the model output parameters, asphalt oil source identification models are constructed based on support vector machine, artificial neural network and decision tree respectively, and the hyper-parameters of the above three algorithms are optimized based on grid search method and cross-validation method to determine the best hyper-parameter combination for each model.

[0059] The optimal parameter results of each model are as follows: Support vector machine: regularization parameter is 10, kernel function is poly, gamma parameter is scale; artificial neural network: activation function is relu, hidden layer size is 30, maximum number of iterations is 1000; decision tree: CART is 4.

[0060] Step 5: Based on accuracy, precision, recall, F1 value and Kappa coefficient, the actual effects of the three asphalt oil source identification models constructed in step 4 under the optimal parameters are evaluated. The index score comparison is as follows: Figure 4 shown.

[0061] Depend on Figure 4 It can be seen that in terms of model evaluation performance, the support vector machine model is inferior to the other two models in all indicators, so it is excluded first. Although the artificial neural network and decision tree perform the same in terms of accuracy, recall and Kappa coefficient, the F1 value of the artificial neural network is slightly higher, indicating that when considering the precision and recall rate, the model has a better balance performance. In summary, linear discriminant analysis is selected as the data set feature extraction method, and the classifier model of the artificial neural network is selected as the asphalt oil source identification model, and a visualization interface is constructed based on this.

[0062] Step 6: Build a visual operation platform for the asphalt oil source identification model based on the PyQT5 tool, and integrate the best feature extraction method of the asphalt dataset determined in step 3 and the best asphalt oil source identification model determined in step 5 into the platform, such as Figure 5After entering all the infrared spectrum characteristic parameters of asphalt, click the confirmation button, and the optimal feature extraction method of the asphalt data set determined in step 3 can be called to automatically extract the infrared spectrum features; then the optimal asphalt oil source identification model determined in step 5 is called and the asphalt oil source is automatically output according to the extracted features, so as to realize the automatic identification of the asphalt oil source.

[0063] In summary, the asphalt oil source identification model and automation method based on the present invention realize the accurate establishment and automatic prediction of the asphalt oil source identification model, and can automatically extract features and automatically output the asphalt oil source based on the input asphalt infrared spectrum parameters. This method of first establishing the asphalt oil source identification model based on machine learning and then visualizing the identification model effectively accelerates the speed and accuracy of asphalt oil source identification and promotes the development of the asphalt material quality identification industry.

Claims

1. A method for automatically implementing an asphalt oil source identification model based on machine learning, characterized in that: The method comprises the following steps: Step 1: Collect asphalt samples from different oil sources, test the infrared spectrum of asphalt, select the characteristic peak height and peak area of ​​infrared spectrum as features, and preliminarily construct an asphalt data set with one-to-one correspondence between infrared spectrum features and oil sources; Step 2: Select the SMOTE algorithm as the data augmentation method for expanding the asphalt dataset. Based on the data augmentation method, expand the number of samples in the original asphalt dataset to optimize the dataset structure and obtain the final asphalt dataset. Step 3: Based on the idea of ​​feature dimensionality reduction, a feature extraction method suitable for asphalt infrared spectrum is studied. The feature extraction method is used to process the features of asphalt infrared spectrum to reduce the number of features, determine the final input parameters of the asphalt identification model, and ensure the performance of the identification model; the feature extraction method suitable for asphalt infrared spectrum based on the idea of ​​feature dimensionality reduction is specifically as follows: Step 31: Extract the features of infrared spectrum of asphalt based on principal component analysis; Step 32: Extract the infrared spectrum characteristics of asphalt based on linear discriminant analysis; Step 33: Visualize the feature extraction results of principal component analysis and linear discriminant analysis to determine the best feature extraction method for asphalt infrared spectral characteristics; Step 4: Using the infrared spectrum features determined in step 3 as the model input parameters and the asphalt oil source as the model output parameters, asphalt oil source identification models are constructed based on support vector machine, artificial neural network and decision tree respectively, and the above three algorithms are optimized based on grid search method and cross validation method to determine the best hyperparameter combination for each model; Step 5: Based on the accuracy, precision, recall, F1 value and Kappa coefficient, the actual effects of the three asphalt oil source identification models constructed in step 4 are evaluated respectively, and the best asphalt oil source identification model is comprehensively determined; Step 6: Build a visual operation platform for the asphalt oil source identification model based on the PyQT5 tool, and integrate the optimal feature extraction method of the asphalt data set determined in step 3 and the optimal asphalt oil source identification model determined in step 5 into the platform to realize the automatic identification of asphalt oil source by inputting asphalt infrared spectral characteristic parameters.

2. The method for automatically implementing the asphalt oil source identification model based on machine learning according to claim 1, characterized in that: In the step 31, the principle of the principal component analysis method is to map high-dimensional data to low-dimensional space while retaining the main features and structure of the data; in the step 32, the principle of the linear discriminant analysis method is to maximize the separability between different categories by finding linear combinations in the feature space; in the step 33, the visualization comparison method is to apply the KMeans clustering algorithm to the data after dimensionality reduction by principal component analysis and linear discriminant analysis, and set the same number of clusters to compare the clustering effects of the two dimensionality reduction methods under the same conditions.

3. The method for automatically implementing the asphalt oil source identification model based on machine learning according to claim 1, characterized in that: In step 4, the basic idea of ​​the support vector machine is to find an optimal hyperplane so that the distance from the data points on both sides of the hyperplane to the hyperplane is maximized; the artificial neural network is a network structure composed of multiple neuron layers, which generates output by receiving input signals and performing weighted summation, and processing them with an activation function; the basic idea of ​​the decision tree is to make decisions by recursively dividing the feature space and constructing a tree; the main idea of ​​the grid search method and the cross-validation method is to exhaustively search all possible parameter combinations and evaluate the performance of each parameter combination through cross-validation, which systematically explores the parameter space and automatically finds the best parameter combination.

4. The method for automatically implementing the asphalt oil source identification model based on machine learning according to claim 1, characterized in that: In the step five, the accuracy rate refers to the ratio of the number of samples correctly predicted by the model to the total number of samples; the precision rate refers to the ratio of all samples predicted to be positive to those that are actually positive. The higher the precision rate, the fewer errors the model makes when predicting positive classes; the recall rate refers to the ratio of all samples that are actually positive to those that are correctly predicted to be positive. The higher the recall rate, the more positive samples the model can identify; the F1 value is the harmonic mean of the precision rate and the recall rate. When both indicators are high, the F1 value will also be high; the Kappa coefficient is used to measure the consistency between the classification results of the classifier and the random classification results. The value range of the Kappa coefficient is [-1, 1]. The closer it is to 1, the more consistent the classification results of the model are with the actual labels.

5. The method for automatically implementing the asphalt oil source identification model based on machine learning according to claim 1, characterized in that: In the step six, first input all the asphalt infrared spectrum characteristic parameters; secondly, click the confirmation button to call the best feature extraction method of the asphalt data set determined in step three to automatically extract the infrared spectrum features; finally, call the best asphalt oil source identification model determined in step five and automatically output the asphalt oil source according to the features after feature extraction, so as to realize the automatic identification of the asphalt oil source.

Images