Basic rock temperature prediction method based on extreme random forest algorithm

Abstract

The invention discloses a basic rock temperature prediction method based on an extreme random forest algorithm, and the method comprises the steps: collecting basic magma experimental data, constructing a database containing the experimental temperature and mineral-melt components under a monohectorite-melt system, and removing unbalanced and extreme experimental data in the database according to preset conditions, the method comprises the following steps: obtaining an original data set, carrying out data preprocessing, obtaining an effective data set, constructing a temperature prediction model based on a limit random forest algorithm, carrying out preliminary training on the model by using the effective data set, and determining an optimal feature combination and a hyper-parameter combination in combination with combined feature selection and hyper-parameter tuning to obtain a final temperature prediction model. And performing temperature prediction. According to the method, the temperature prediction model which does not depend on the water content as characteristic input is constructed, the technical problem of measuring the water content in natural minerals and rocks is effectively avoided, and the method does not depend on prior thermodynamic hypothesis and has good adaptability and universality.

Description

Technical Field

[0001] The present invention relates to the field of geological temperature calculation science and technology, and in particular to a basic rock temperature prediction method based on an extreme random forest algorithm. Background Art

[0002] The temperature at the time of mafic magma formation is a key parameter for understanding mantle-crust interaction mechanisms, inferring plate tectonic settings, and reconstructing the deep Earth's thermodynamic environment. Empirical equations based on mineral-melt chemical equilibrium, developed through experimental calibration and thermodynamic modeling, are the current mainstream technology for predicting mafic rock temperatures. However, these equations often employ linear or low-order polynomial regression fits, which struggle to accurately capture the complex nonlinear mapping relationship between compositional data and temperature. Their prediction accuracy is limited, and their application is typically restricted to specific compositional systems. With the continued expansion of artificial intelligence in Earth sciences, some studies have attempted to incorporate machine learning methods to construct temperature prediction models. However, these studies have not performed systematic feature screening and have primarily relied on water in melts as an input variable. Because the precise determination of water content in natural rocks and minerals is difficult, and estimating water content through indirect empirical formulas is prone to multiple sources of error, these models still face certain limitations in practical application.

[0003] A basic rock temperature prediction method based on the extreme random forest algorithm realizes the prediction of basic rocks by integrating machine learning by constructing an extreme random forest model that does not use water content as a feature training. It can automatically mine the implicit rules in high-dimensional data, eliminating the prior assumptions of traditional empirical equations, and providing new ideas for solving the inversion problem of complex diagenetic physical conditions. Summary of the Invention

[0004] The purpose of the present invention is to provide a basic rock temperature prediction method based on extreme random forest algorithm.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions: The present invention provides a basic rock temperature prediction method based on an extreme random forest algorithm, comprising: S1 collects basic magma experimental data and constructs a database containing experimental temperature and mineral-melt composition in the clinopyroxene-melt system; S2 removes the non-equilibrium and extreme experimental data in the database according to preset conditions to obtain the original data set; S3 performs data preprocessing on the original data set to obtain a valid data set; the valid data set includes melt characteristics, clinopyroxene characteristics, and derived characteristics; the derived characteristics include the magnesium index of clinopyroxene and the magnesium-iron partition coefficient of clinopyroxene-melt; S4 builds a temperature prediction model based on the extreme random forest algorithm, uses a valid data set to perform preliminary training on the model, and combines combined feature selection and hyperparameter tuning to determine the optimal feature combination and hyperparameter combination to obtain the final temperature prediction model; S5 performs temperature prediction based on the final temperature prediction model.

[0006] Preferably, the preset conditions in S2 include experimental temperature threshold, component content range and mineral-melt magnesium-iron partition coefficient restrictions.

[0007] Preferably, the data preprocessing method described in S3 includes: performing multivariate regression interpolation on missing values ​​in the original data set, and selecting melt characteristics and monoclinic pyroxene characteristics from the processed original data set, further constructing derived characteristics, performing central logarithmic transformation on the composition data in the characteristics, and then performing Z-score standardization.

[0008] Preferably, the melt characteristics in S3 include SiO2, TiO2, Al2O3, FeO, MgO, MnO, CaO, Na2O and K2O characteristics of the melt, and the clinopyroxene characteristics include SiO2, TiO2, Al2O3, Cr2O3, FeO, MgO, MnO, CaO and Na2O characteristics of clinopyroxene.

[0009] Preferably, the method steps for training the model using a valid data set in S4 are specifically: The effective data set is divided into equal frequency according to the variable temperature to obtain several data subsets; The same proportion of samples were extracted from each data subset to construct a training set and a validation set, and the temperature prediction model was trained based on the training set.

[0010] Preferably, the method steps described in S4 and combined with combined feature selection and hyperparameter tuning to determine the optimal feature combination and hyperparameter combination are specifically: S41 combines filtering feature selection and forward feature selection to determine the optimal feature combination: Filter feature selection: The temperature prediction model is trained a preset number of times, the average contribution of different features to temperature prediction is extracted, and the top n features are selected to construct the initial feature subset; Forward feature selection: Based on the initial feature subset, introduce features that can improve model performance and build the optimal feature combination until the introduced features can no longer improve model performance; S42 performs a grid search on the number of decision trees and the maximum number of features for node partitioning in the temperature prediction model to generate a set of hyperparameter combinations. Based on the optimal feature combination in S41, the performance of the model under different hyperparameter combinations is evaluated, and the optimal hyperparameter combination is selected.

[0011] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) By constructing a characteristic input system that does not introduce water content, the present invention breaks through the existing technology's reliance on the determination of specific magma types and water content parameters, and significantly improves the accuracy and applicability of temperature prediction.

[0012] (2) In terms of feature construction, the present invention not only uses the main oxide components of melts and clinopyroxene, but also introduces geochemical derived features to enhance the model's ability to express temperature-sensitive features.

[0013] (3) The present invention optimizes the temperature prediction model through a combination of hyperparameter optimization and feature importance ranking to enhance the prediction accuracy and robustness of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of a basic rock temperature prediction method based on the extreme random forest algorithm of the present invention; Figure 2 Schematic diagram of evaluation results of different machine learning algorithms in an embodiment of the present invention; Figure 3 A bar chart showing the importance of each feature in an embodiment of the present invention; Figure 4 Schematic diagram of forward feature selection results based on five-fold cross validation in an embodiment of the present invention; Figure 5 1 is a diagram of the residual-frequency distribution and the true value-predicted value of the temperature prediction model in an embodiment of the present invention based on an independent validation set. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0016] Reference Figure 1 As shown, the present invention provides a basic rock temperature prediction method based on the extreme random forest algorithm, comprising the following steps: S1 collects basic magma experimental data and constructs a database containing experimental temperature and mineral-melt composition in the monoclinic pyroxene-melt system.

[0017] In this example, basic rock data from mainstream domestic and foreign journals and global open source databases were sorted to construct a database containing experimental temperatures and mineral-melt compositions in the monoclinic pyroxene-melt system.

[0018] S2 removes non-equilibrium and extreme experimental data from the database according to preset conditions to obtain the original data set.

[0019] In this embodiment, the preset conditions are as follows: the experimental temperature must be less than 1800 , the SiO2 content of the melt must be greater than 35%, the Mg-Fe partition coefficient of the mineral-melt is the average of all samples Within the range of times the standard deviation, a total of 2422 sample records were obtained.

[0020] S3 performs data preprocessing on the original data set to obtain a valid data set; the valid data set includes melt characteristics, clinopyroxene characteristics, and derived characteristics; the derived characteristics include the magnesium index of clinopyroxene and the magnesium-iron partition coefficient of clinopyroxene-melt; In some embodiments, the data preprocessing method described in S3 includes: performing multivariate regression interpolation on missing values ​​in the original data set, and selecting melt characteristics and monoclinic pyroxene characteristics from the processed original data set, further constructing derived characteristics, performing a central logarithmic transformation on the composition data in the characteristics, and then performing Z-score standardization.

[0021] Among them, the melt characteristics include SiO2, TiO2, Al2O3, FeO, MgO, MnO, CaO, Na2O and K2O characteristics of the melt, and the monoclinic pyroxene characteristics include SiO2, TiO2, Al2O3, Cr2O3, FeO, MgO, MnO, CaO and Na2O characteristics of monoclinic pyroxene.

[0022] It needs to be explained that the introduction of geochemical derived characteristics can improve the temperature fitting ability of the temperature prediction model and help to characterize the thermodynamic response characteristics between minerals and melts.

[0023] S4 builds a temperature prediction model based on the extreme random forest algorithm, uses a valid data set to perform preliminary training on the model, and combines combined feature selection and hyperparameter tuning to determine the optimal feature combination and hyperparameter combination to obtain the final temperature prediction model.

[0024] Specifically, such as Figure 2 As shown in the figure, the performance of four mainstream machine learning algorithms, namely multivariate linear regression, extreme random forest, Gaussian process regression and support vector regression, is evaluated. The training set is divided into training subsets and test sets in a ratio of 4:1, and repeated 100 times to reduce the randomness caused by a single division. The coefficient of determination R of different models in each round of training is recorded.2 and standard error SEE, and finally the optimal extreme random forest algorithm is selected to build the model.

[0025] In this embodiment, the valid data set is divided into 10 data subsets based on the variable temperature. 20% of the samples are extracted from each data subset to construct a validation set, and the remaining samples are combined into a training set. The validation set contains a total of 485 samples, and the training set contains a total of 1937 samples. The temperature prediction model is trained based on the training set.

[0026] It should be explained that in the original data set, the equal-frequency binning method is used based on the temperature variable to divide the samples into several subsets, which can ensure the balance of the number of samples in each subset. At the same time, in each subset, the training set and validation set are divided into equal proportions. By controlling the sampling ratio in each subset, the uniformity of the validation set within the temperature distribution range is guaranteed, which improves the accuracy and reliability of the evaluation of the model's generalization ability under different temperature conditions.

[0027] In this embodiment, the average contribution of different features to temperature prediction in the extreme random forest model is extracted, and the following is obtained: Figure 3 The importance of each feature is shown in the figure, and the top 8 features are selected to construct the initial feature subset, including: SiO2, MgO, CaO and FeO of melt, FeO and CaO of clinopyroxene, magnesium index of clinopyroxene, and magnesium-iron partition coefficient of clinopyroxene-melt.

[0028] like Figure 4 As shown in the figure, the performance of the model changes with the number of features. When the number of features is 16, the performance of the model is the best. In addition to the 8 features in the initial subset, the optimal feature combination also includes TiO2, Al2O3, Na2O, K2O, MnO of the melt and TiO2, Al2O3, Na2O features of monoclinic pyroxene.

[0029] Specifically, by using a combined feature selection strategy, we avoid the shortcomings of forward feature selection that easily falls into local optimality and the filtering method that usually ignores the complex interactions between features. This helps to screen out more globally optimal feature combinations, thereby improving model performance.

[0030] In this example, a grid search was performed on the number of decision trees and the number of node partition features to generate a set of hyperparameter combinations. All hyperparameter combinations were evaluated through 100 five-fold cross validations. Finally, 600 decision trees and 15 maximum node partition features were selected as the optimal hyperparameter combination. The coefficient of determination and standard error of the model under different hyperparameter combinations are shown in Tables 1 and 2: Table 1 Average values ​​of coefficient of determination for different hyperparameter combinations

[0031] Table 2. Mean standard errors of different hyperparameter combinations

[0032] In this embodiment, the final temperature prediction model is obtained based on the optimal feature combination and hyperparameter combination, and the model file is saved. The temperature prediction model has a significant fitting effect with a determination coefficient of 0.938 and a standard error of 38.01 on the training set.

[0033] S5 uses the trained temperature prediction model to predict the temperature of basic rocks; In this embodiment, if Figure 5 As shown in the figure, the final temperature prediction model showed significant goodness of fit and prediction stability on the validation set, with a determination coefficient of 0.934 and a standard error of 38.95. There was no overfitting, and the final temperature prediction model had good versatility.

[0034] The above content is merely an example and explanation of the structure of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the scope of protection of the present invention.

Claims

1. A basic rock temperature prediction method based on extreme random forest algorithm, characterized in that: The following steps are involved: S1 collects basic magma experimental data and constructs a database containing experimental temperature and mineral-melt composition in the clinopyroxene-melt system; S2 removes the non-equilibrium and extreme experimental data in the database according to preset conditions to obtain the original data set; S3 performs data preprocessing on the original data set to obtain a valid data set; the valid data set includes melt characteristics, clinopyroxene characteristics, and derived characteristics; the derived characteristics include the magnesium index of clinopyroxene and the magnesium-iron partition coefficient of clinopyroxene-melt; S4 builds a temperature prediction model based on the extreme random forest algorithm, uses a valid data set to perform preliminary training on the model, and combines combined feature selection and hyperparameter tuning to determine the optimal feature combination and hyperparameter combination to obtain the final temperature prediction model; S5 performs temperature prediction based on the final temperature prediction model.

2. The method for predicting basic rock temperature based on the extreme random forest algorithm according to claim 1, characterized in that: The preset conditions described in S2 include the experimental temperature threshold, component content range, and the limitation of mineral-melt Mg-Fe partition coefficient.

3. The method for predicting basic rock temperature based on the extreme random forest algorithm according to claim 1, characterized in that: The data preprocessing method described in S3 includes: performing multivariate regression interpolation on missing values ​​in the original data set, selecting melt characteristics and monoclinic pyroxene characteristics from the processed original data set, further constructing derived characteristics, performing central logarithmic transformation on the composition data in the characteristics, and then performing Z-score standardization.

4. The method for predicting basic rock temperature based on the extreme random forest algorithm according to claim 1, characterized in that: The melt characteristics described in S3 include SiO2, TiO2, Al2O3, FeO, MgO, MnO, CaO, Na2O and K2O characteristics of the melt, and the monoclinic pyroxene characteristics include SiO2, TiO2, Al2O3, Cr2O3, FeO, MgO, MnO, CaO and Na2O characteristics of monoclinic pyroxene.

5. The method for predicting basic rock temperature based on extreme random forest algorithm according to claim 1, characterized in that: The method steps for training the model using a valid dataset described in S4 are specifically as follows: The effective data set is divided into equal frequency according to the variable temperature to obtain several data subsets; The same proportion of samples were extracted from each data subset to construct a training set and a validation set, and the temperature prediction model was trained based on the training set.

6. The method for predicting basic rock temperature based on extreme random forest algorithm according to claim 1, characterized in that: The method steps described in S4 and combined with combined feature selection and hyperparameter tuning to determine the optimal feature combination and hyperparameter combination are as follows: S41 combines filtering feature selection and forward feature selection to determine the optimal feature combination: Filter feature selection: The temperature prediction model is trained a preset number of times, the average contribution of different features to temperature prediction is extracted, and the top n features are selected to construct the initial feature subset; Forward feature selection: Based on the initial feature subset, introduce features that can improve model performance and build the optimal feature combination until the introduced features can no longer improve model performance; S42 performs a grid search on the number of decision trees and the maximum number of features for node partitioning in the temperature prediction model to generate a set of hyperparameter combinations. Based on the optimal feature combination in S41, the performance of the model under different hyperparameter combinations is evaluated, and the optimal hyperparameter combination is selected.

Images