Kernel Support Vector Regression Research Articles

Electric Vehicle Charging Load Forecasting Based on K-Means++-GRU-KSVR

The accurate short-term forecasting of an electric vehicle (EV) load is crucial for the reliable operation of a power grid and for effectively reducing energy consumption. Due to the fluctuations in EV charging loads, particularly the significant load variation between commercial and non-commercial areas, global models often suffer from prediction errors when forecasting loads. To address this issue, this paper proposes a regional forecasting method based on K-means++ clustering and deep learning algorithms. First, the K-means++ algorithm was used to partition the data into different regions, and an independent load-forecasting model was established for each region. Then, a combination of kernel support vector regression (KSVR) and gated recurrent unit (GRU) models was used to handle nonlinear features and time-dependent data, where particle swarm optimization (PSO) further optimized the model parameters to improve the forecasting accuracy. Finally, a weighted summation method was used to integrate the forecast results from each region, resulting in a more accurate overall load forecast. The experimental results show that the proposed model provided better prediction performance by capturing the spatiotemporal characteristics of the EV charging load, effectively addressing the challenges posed by regional differences, and outperforming the single-model forecasts.

Machine learning complete intersection Calabi-Yau 3-folds

Gaussian process regression, kernel support vector regression, the random forest, extreme gradient boosting, and the generalized linear model algorithms are applied to data of complete intersection Calabi-Yau threefolds. It is shown that Gaussian process regression is the most suitable for learning the Hodge number h2,1 in terms of h1,1. The performance of this regression algorithm is such that the Pearson correlation coefficient for the validation set is R2=0.9999999995 with a root mean square error RMSE=0.0002895011. As for the train set, these two parameters are as follows: R2=0.9999999994 and RMSE=0.0002854348. The training error and the cross-validation error of this regression are 1×10−9 and 1.28×10−7, respectively. Learning the Hodge number h1,1 in terms of h2,1 yields R2=1.000000 and RMSE=7.395731×10−5 for the validation set of the Gaussian process regression. Published by the American Physical Society 2024

Co-training machine learning enables interpretable discovery of near-infrared phosphors with high performance

Near-infrared (NIR) phosphors based on Cr3+ doped garnets present great potential in the next generation of NIR light sources. Nevertheless, the huge searching space for the garnet composition makes the rapid discovery of NIR phosphors with high performance remain a great challenge for the scientific community. Herein, a generalizable machine learning (ML) strategy is designed to accelerate the exploration of innovative NIR phosphors via establishing the relationship between key parameters and emission peak wavelength (EPW). We propose a semi-supervised co-training model based on kernel ridge regression (KRR) and support vector regression (SVR), which successfully establishes an expanded dataset with unlabeled dataset (previously unidentified garnets), addressing the overfitting issue resulted from a small dataset and greatly improving the model generalization capability. The model is then interpreted to extract valuable insights into the contribution originated from different features. And a new type NIR luminescent material of Lu3Y2Ga3O12: Cr3+ (EPW~750 nm) is efficiently screened, which demonstrates a high internal (external) quantum efficiency of 97.1% (38.8%) and good thermal stability, particularly exhibiting promising application in the NIR phosphor-converted LEDs (pc-LED). These results suggest the strategy proposed in this work could provide new viewpoint and direction for developing NIR luminescence materials.

Machine learning based prediction of flyrock distance in rock blasting: A safe and sustainable mining approach

Flyrock is a significant environmental and safety concern in mining and construction. It arises from various geological and blast design factors, posing risks to workers, machinery, and nearby structures. This study examined how these factors affect the rate and distance of flyrock projections caused by blasts. To address this issue, advanced machine learning (ML) models were used to predict flyrock distances in the Akoko Edo dolomite quarries. The models examined included bidirectional recurrent neural networks (BRNNs), support vector regression (SVR) with different kernels (SVR-S, SVR-RBF, SVR-L, SVR-P), long short-term memory (LSTM) networks, and random forest (RF) algorithms. A case study was conducted using 258 blasting data samples to develop these models. Key factors influencing flyrock were identified: blast hole burden distance, maximum instantaneous charge, and rock brittleness index. Using these factors, a flyrock possibility assessment chart was created to enhance the safety of small-scale mining operations. The model’s prediction accuracy was evaluated using correlation coefficients and four performance metrics. The LSTM model stood out, achieving the highest coefficient of correlation (R2 = 0.99) for both training and testing datasets. This indicates that the LSTM model accurately predicts blast-induced flyrock distance. The study also revealed that the Gaussian-RBF kernel SVR has high prediction accuracy when compared to other SVR variants (SVR-S, SVR-L, and SVR-P). In conclusion, the study compared various ML models for flyrock reduction and found that the LSTM model was the most effective in estimating blast-induced flyrock distances.

Inverse finite element method and support vector regression for automated crack detection with OFDR-Distributed fiber optic sensors

Crack detection is a crucial but challenging task. In this paper, a new method is introduced for automatical detection of the substrate cracks. An anomaly index is defined based on optical frequency domain reflectometry (OFDR) measured strain and inverse finite element method (iFEM) recovered strain. The index is sensitive to crack-induced strain but not sensitive to substrate strain. Leveraging on the self-defined anomaly index, the crack location can be identified with millimetre accuracy. Then, the detection of the crack depth is regarded as a typical multi-regression analysis and solved by support vector regression (SVR) method. Particle swarm optimization (PSO) approach is applied to find optimal parameters of the SVR kernel function. Experimental cases of a simple-supported beams with different crack locations and depths are performed, demonstrating the excellent prediction accuracy of the method. The proposed method has a promising potential in identification and quantification of cracks.

A materials informatics framework based on reduced‐order models for extracting structure–property linkages of additively manufactured continuous fiber‐reinforced polymer composites

AbstractThe innovative combination of additive manufacturing (AM) and continuous fiber‐reinforced polymer composites (CFRPCs) confers products with the dual advantages of integrated manufacturing and designability of properties, but lack an efficient and reliable method for property prediction. This study presents a materials informatics framework using reduced‐order models and machine learning (ML) to extract the structure–property (SP) linkages between microstructures and macroscopic elastic properties of AM‐CFRPCs. The initial step involves generating microstructural 2D cross‐sections and representative volume elements (RVEs) with random fiber and pore distributions based on the minimum potential method. Then, finite element (FE) calculations are performed on RVEs to obtain nine macroscopic elastic properties. Following that, the quantification and dimensionality reduction of the 2D cross‐sectional dataset are conducted separately using two‐point spatial correlations and principal component analysis (PCA). Finally, a Bayesian optimized composite kernel support vector regression (CK‐SVR) algorithm is used to effectively establish complex mapping relationships between the reduced‐order representations of the microstructures and the mechanical properties. Despite the reduced‐order dataset containing only 3–6 variables, the framework generates an interpretable model exhibiting excellent accuracy with all predicted R2 values surpassing 0.91. Therefore, this framework presents a prospective solution for expediting the design and optimization of AM‐CFRPCs.Highlights A materials informatics scheme is proposed to predict the 9 elastic properties of AM‐CFRPCs. Microstructures are quantified and dimensionally reduced by two‐point statistics and PCA. SP linkages are established between 2D cross‐sections and 3D macromechanical properties. Modified CK‐SVR exhibits higher prediction accuracy compared to conventional models.

Advancing SDGs: Predicting Future Shifts in Saudi Arabia’s Terrestrial Water Storage Using Multi-Step-Ahead Machine Learning Based on GRACE Data

The availability of water is crucial for the growth and sustainability of human development. The effective management of water resources is essential due to their renewable nature and their critical role in ensuring food security and water safety. In this study, the multi-step-ahead modeling approach of the Gravity Recovery and Climate Experiment (GRACE) terrestrial water storage (TWS) was utilized to gain insights into and forecast the fluctuations in water resources within Saudi Arabia. This study was conducted using mascon solutions obtained from the University of Texas Center for Space Research (UT-CSR) over the period of 2007 to 2017. The data were used in the development of artificial intelligence models, namely, an Elman neural network (ENN), a backpropagation neural network (BPNN), and kernel support vector regression (k-SVR). These models were constructed using various input variables, such as t-12, t-24, t-36, t-48, and TWS, with the output variable being the focus. A simple and weighted average ensemble was introduced to improve the accuracy of marginal and weak predictive results. The performance of the models was assessed with the use of several evaluation metrics, including mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), correlation coefficient (CC), and Nash–Sutcliffe efficiency (NSE). The results of the estimate indicate that k-SVR-M1 (NSE = 0.993, MAE = 0.0346) produced favorable outcomes, whereas ENN-M3 (NSE = 0.6586, MAE = 0.6895) emerged as the second most effective model. The combinations of all other models exhibited accuracies ranging from excellent to marginal, rendering them unreliable for decision-making purposes. Error ensemble methods improved the standalone model and proved merit. The results also serve as an important tool for monitoring changes in global water resources, aiding in drought management, and understanding the Earth’s water cycle.

Research on the Liquor Yield Prediction Model Integrating Hybrid Kernel Support Vector Regression and Dung Beetle Optimizer

Research on the Liquor Yield Prediction Model Integrating Hybrid Kernel Support Vector Regression and Dung Beetle Optimizer

AI-based Optimization of Tensile Strength of the Cement Concrete Incorporating Recycled Mixed Plastic Fine used in Road Construction

Abstract: One of the main problems in materials science and engineering is predicting the tensile strength of materials. In this study, we investigate how to model and forecast tensile strength (Tensile Strength in Mpa) based on different material attributes using Support Vector Regression (SVR) using Linear and Polynomial Kernels. The dataset includes the following details: plastic type, fine aggregate ratio, water/cement ratio, cement content, and associated tensile strength values. This work has two main goals: (1) to assess the predictive power of SVR models with various kernel functions and (2) to examine the significance of unique material attributes for prediction. To simulate the link between the input features and tensile strength, we used SVR in conjunction with a Linear Kernel. The final model included insightful information on how each feature affected the forecast. Our results show that the Polynomial Kernel SVR model may better reflect the complex interactions among the material attributes than the Linear Kernel SVR model, despite being more interpretable. Better prediction performance was offered by the Polynomial Kernel SVR, which also revealed the non-linear dependencies in the data.

Prediksi Jumlah Perceraian Menggunakan Metode Support Vector Regression (SVR)

The increasing number of divorces poses an increasingly significant social challenge in Indonesia, including in the city of Pekanbaru. The impact of these divorces on the adolescent population can have negative effects on their emotional and psychological well-being, as well as their ability to interact socially and engage in the learning process. This study utilizes monthly divorce data from 2015 to April 2023 to conduct time series analysis and applies the Support Vector Regression (SVR) method to predict the number of divorces in the city of Pekanbaru. Three types of SVR kernels, namely linear, polynomial, and radial basis function (RBF), are evaluated and compared to find the kernel with the best Mean Squared Error (MSE) results. Through grid search analysis, optimal parameter values for each kernel are determined. The test results indicate that the SVR model with a polynomial kernel provides more accurate predictions with an MSE of 0.010228, compared to the linear kernel (MSE = 0.012767) and the RBF kernel (MSE = 0.010812).

A Framework for Credit Risk Prediction Using the Optimized-FKSVR Machine Learning Classifier

Transparency is influenced by several crucial factors, such as credit risk (CR) predictions, model reliability, efficient loan processing, etc. The emergence of machine learning (ML) techniques provides a promising solution to address these challenges. However, it is the responsibility of banking or nonbanking organizations to control their approach to incorporate this innovative methodology to mitigate human preferences in loan decision-making. The research article presents the Optimized-Feature based Kernel Support Vector Regression (O-FKSVR) model which is an ML-based CR analysis model in the digital banking. This proposal aims to compare several ML methods to identify a precise model for CR assessment using real credit database information. The goal is to introduce a classification model that uses a hybrid of Stochastic Gradient Descent (SGD) and firefly optimization (FFO) methods with Support Vector Regression (SVR) to predict credit risks in the form of probability, loss given, and exposure at defaults. The proposed O-FKSVR model extracts features and predicts outcomes based on data gathered from online credit analysis. The proposed O-FKSVR model has increased the accuracy rate and resolved the existing problems. The experimental study is conducted in Python, and the results demonstrate improvements in accuracy, precision, and reduced error rates compared to previous ML methods. The proposed O-FKSVR model has achieved a maximum accuracy rate value of 0.955%, precision value of 0.96%, and recall value of 0.952%, error rate value of 4.4 when compared with the existing models such as SVR, DT, RF, and AdaBoost.

Urban river ammonia nitrogen prediction model based on improved whale optimization support vector regression mixed synchronous compression wavelet transform

Accurate prediction of ammonia nitrogen (NH3–N) in water was crucial to efficient and reasonable judgment of water quality and environmental conditions. A model named SCWT-IWOA-SVR was proposed in this study by combining synchronous compressed wavelet transform (SCWT), improved whale optimization algorithm (IWOA) and support vector regression (SVR). Firstly, the SCWT algorithm was used to de-noise the original data, overcome the nonlinearity of the data, and explore the variation rule of the water quality index time series. Statistical methods and Pearson correlation analysis were then used to select features. IWOA was introduced to optimize the hyperparameters in SVR kernel function, tent mapping was used to randomly initialize the location of the whales, making them more evenly distributed, combined wheel selection mechanism was used to select more representative prey, which leading to improve the accuracy of forecasts results. The results indicated that the SCWT-IWOA-SVR model, which incorporates synchronous compression wavelet transform and the improved whale optimization algorithm, exhibit a better prediction effect compared with SVR, WOA-SVR, WOA-SVR, and Random Forest (RF) models. The results indicated that the model has a good prediction performance of low error and high generalization (RMSE = 0.055, MAE = 0.015, R2 = 0.989), the RMSE <0.2 and R2 > 0.915 of the proposed SCWT-IWOA-SVR model in three days advance predicted. SCWT-IWOA-SVR model can provide a scientific basis for the comprehensive treatment and decision-making of water pollutants in the watershed.

Predicting narcissistic personality traits from brain and psychological features: A supervised machine learning approach

ABSTRACTNarcissism is a multifaceted construct often linked to pathological conditions whose neural correlates are still poorly understood. Previous studies have reported inconsistent findings related to the neural underpinnings of narcissism, probably due to methodological limitations such as the low number of participants or the use of mass univariate methods. The present study aimed to overcome the previous methodological limitations and to build a predictive model of narcissistic traits based on neural and psychological features. In this respect, two machine learning-based methods (Kernel Ridge Regression and Support Vector Regression) were used to predict narcissistic traits from brain structural organization and from other relevant normal and abnormal personality features. Results showed that a circuit including the lateral and middle frontal gyri, the angular gyrus, Rolandic operculum, and Heschl’s gyrus successfully predicted narcissistic personality traits (p < 0.003). Moreover, narcissistic traits were predicted by normal (openness, agreeableness, conscientiousness) and abnormal (borderline, antisocial, insecure, addicted, negativistic, machiavellianism) personality traits. This study is the first to predict narcissistic personality traits via a supervised machine learning approach. As such, these results may expand the possibility of deriving personality traits from neural and psychological features.

IMPLEMENTASI GRIDSEARCHCV PADA SUPPORT VECTOR REGRESSION (SVR) UNTUK PERAMALAN HARGA SAHAM

Stock is a sign of the capital participation of a person or authority in a company (PT). PT Anabatic Technologies Tbk (ATIC) is one of the service providers and IT consultants that is included in the technology sector, which is a new sector in the IDX-IC classification. ATIC stock trading was temporarily suspended due to a significant increase in cumulative prices. This indicates that stock prices tend to be volatile and non-linear. The Support Vector Regression (SVR) method can be used to predict stock prices. SVR is able to solve non-linear data problems by using kernel functions so it can overcome overfitting problems and will give good performance. The SVR problem is difficult to determine the optimal hyperparameters, so this research uses grid search cross validation (GridSearchCV). In this research, ATIC’s daily closing price data was used with 1007 training data and 252 testing data. The results show that the best model is SVR with a linear kernel and the hyperparameters used are Cost and epsilon . The linear kernel SVR model produces a MSE of 0,001237173; SMAPE of 0,1167301; and = 0,9206643

Assessment of rolling element bearing degradation based on Dynamic Time Warping, kernel ridge regression and support vector regression

An accurate and reliable forecast for fault/degradation trend has an enormous significance in predictive maintenance of industrial systems. Prognosis of bearing is essential for efficient implementation of condition based maintenance of rotating machinery to prevent sudden/unexpected breakdown. The Dynamic Time Warping (DTW) algorithm is proposed for the first time for bearing prognosis assessment and health trend prediction of rolling element bearing. The Dynamic Time Warping based value is then applied to obtain a Confidence Value (CV) to standardize the degradation process. The range of the parameter CV is zero to unity. The various degradation stages are recognized by using CV and the fault frequencies are obtained from the envelop spectrum. Further, the Kernel Ridge Regression (KRR) and Support Vector Regression (SVR) methodologies are applied on the DTW feature to forecast the bearing degradation trend. The effectiveness of DTW-KRR and DTW-SVR is analyzed by various accuracy metrics. The present study shows that the DTW and DTW-SVR are effective tools for bearing health degradation assessment and prognosis of rolling element bearing.

Prediction of PV cell parameters at different temperatures via ML algorithms and comparative performance analysis in Multiphysics environment

As the world moves towards renewable resources to meet the heaping power demand, the dependence on Photo Voltaic (PV) panel-based power generation has globally touched 1000TWh in 2021. These panels made up of different materials, such as Germanium (Ge), Silicon (Si), Indium Phosphide (InP), and Gallium Arsenide (GaAs), work in wider applications ranging from earth to space. But in real-time environmental conditions, these materials exhibit only 11 to 15 % efficiency, mainly due to thermal loss and exposure to variable irradiation and temperature conditions. Thus, in the current research, the PV cell/panel is modeled in an experimentally validated Multiphysics environment at temperatures from – 45 0C to + 51 0C using Ge, Si, InP, and GaAs as materials.The efficiency of the PV cell/panel is estimated in the current research based on the thermal losses within the material, the width of the bandgap, and the thickness of the cell. To analyze the thermal losses, joule heat generation of PV cell/panels made of all different materials is obtained. It is observed that at ambient temperature, although the GaAs and InP have an energy bandgap of 1.42 eV and 1.34 eV, respectively, the joule heating effect is minimum (3.95KW/m3 - InP, 5.05KW/m3- GaAs) when compared to Si (61KW/m3) and Ge (234KW/m3). But InP showed lower efficiency due to the thickness of the cell, which prevents the penetration of photons deeper into the inner layers. Further, in this research, the effect of real-time temperature conditions is obtained for all the PV panels by modeling them at −45 °C, 0 °C, 25 °C, and 51 °C. Parameters such as short circuit current (Isc), open circuit voltage (Voc), Fill Factor (FF), and maximum power (Pmax) are enumerated along with the efficiency. For a faster parametric analysis at all possible real-time temperature conditions in Indian climatic scenario, in this research, an ML based prediction model is developed using Kernel Ridge Regression (KRR), Polynomial Regression (PR), Linear Regression (LR), and Support Vector Regression (SVR). It is observed that, the KRR algorithm can give faster (in few seconds), effective, and error-free prediction (92.23% of accuracy). The prediction results are validated with the Multiphysics environment, data sheet, and experimental data.

ANALISIS SUPPORT VECTOR REGRESSION (SVR) DENGAN ALGORITMA GRID SEARCH TIME SERIES CROSS VALIDATION UNTUK PREDIKSI JUMLAH KASUS TERKONFIRMASI COVID-19 DI INDONESIA

Coronavirus Disease 2019 or Covid-19 is a group of types of viruses that interfere with the respiratory tract associated with the seafood market that emerged in Wuhan City, Hubei Province, China at the end of 2019. The first confirmed cases of Covid-19 in Indonesia on March 2, 2020, were 2 cases and until the end of 2021, it continues to grow every day. The purpose of this study was to predict the number of confirmed cases of Covid-19 in Indonesia using the Support Vector Regression (SVR) method with linear kernel functions, radial basis functions (RBF), and polynomials. Support Vector Regression (SVR) is the application of a support vector machine (SVM) in regression cases that aims to find the dividing line in the form of the best regression function. The advantage of the SVR model is can be used on time series data, data that are not normally distributed and data that is not linear. Parameter selection for each kernel used a grid search algorithm combined with time series cross validation. The criteria used to measure the goodness of the model are MSE (Mean Square Error), MAPE (Mean Absolute Percentage Error) and R2 (Coefficient of Determination). The results of this study indicate that the best model is Support Vector Regression (SVR) with a polynomial kernel and the parameters used include Cost = 1, degree = 1, and coefficient = 0.1. The polynomial kernel SVR model produces a MAPE value of 0.4946215%, which means the model has very good predictive ability. The prediction accuracy obtained with an R2 value of 85.65011% and an MSE value of 161606.1.

Predicting oil field performance using machine learning programming: a comparative case study from the UK continental shelf

Predicting the performance of a subsurface oil field is a large, multivariant problem. Production is controlled and influenced by a wide array of geological and engineering parameters which overlap and interact in ways that are difficult to unravel in a manner that can be predictive. Supervised machine learning is a statistical approach which uses empirical learnings from a training dataset to create models and make predictions about future outcomes. The goal of this study is to test a number of supervised machine learning methods on a dataset of oil fields from the United Kingdom continental shelf (UKCS), in order to assess whether, (a) it is possible to predict future oil field performance and (b), which methods are the most effective. The study is based on a dataset of 60 fields with 5 controlling parameters, (gross depositional environment, average permeability, net-to-gross, gas–oil ratio and total number of wells) and 2 outcome parameters (recovery factor and maximum field rate) for each. The choice of controlling parameters was based on a PCA of a larger dataset from a wider project database. Five different machine learning algorithms were tested. These include linear regression, robust linear regression, linear kernel support vector regression, cubic kernel support vector regression and boosted trees regression. Overall, 83% of the data was used as a training dataset while 17% was used to test the predictability of the algorithms. Results were compared using R-Squared, Mean Square Error, Root Mean Square Error and Mean Absolute Error. Graphs of predicted responses v. true (actual) responses are also shown to give a visual illustration of model performance. Results of this analysis show that certain methods perform better than others, depending on the outcome variable in question (recovery factor or maximum field rate). The best method for both outcome variables was the support vector regression, where, depending on the kernel function applied, a reliable level of predictability with low error rates were achieved. This demonstrates a strong potential for statistics-based prediction models of reservoir performance. Thematic collection: This article is part of the Digitally enabled geoscience workflows: unlocking the power of our data collection available at https://www.lyellcollection.org/topic/collections/digitally-enabled-geoscience-workflows

Prediction of band gap of transition metal sulfide with Janus structure by deep learning atomic feature representation method

With the development of artificial intelligence, machine learning (ML) is more and more widely used in material computing. To apply ML to the prediction of material properties, the first thing to do is to obtain effective material feature representation. In this paper, an atomic feature representation method is used to study a low-dimensional, densely distributed atomic eigenvector, which is applied to the band gap prediction in material design. According to the types and numbers of atoms in the chemical formula of material, the Transformer Encoder is used as a model structure, and a large number of material chemical formula data are trained to extract the features of the training elements. Through the clustering analysis of the atomic feature vectors of the main group elements, it is found that the element features can be used to distinguish the element categories. The Principal Component Analysis of the atomic eigenvector of the main group element shows that the projection of the atomic eigenvector on the first principal component reflects the outermost electron number corresponding to the element. It illustrates the effectiveness of atomic eigenvector extracted by using the transformer model. Subsequently, the atomic feature representation method is used to represent the material characteristics. Three ML methods named Random Forest (RF), Kernel Ridge Regression (KRR) and Support Vector Regression (SVR) are used to predict the band gap of the two-dimensional transition metal chalcogenide compound MXY (M represents transition metal, X and Y refer to the different chalcogenide elements) with Janus structure. The hyperparameters of ML model are determined by searching for parameters. To obtain stable results, the ML model is tested by 5-fold cross-validation. The results obtained from the three ML models show that the average absolute error of the prediction using atomic feature vectors based on deep learning is smaller than that obtained from the traditional Magpie method and the Atom2Vec method. For the atomic eigenvector method proposed in this paper, the prediction accuracy of the KRR model is better than that of the results obtained from the Magpie method and Atom2Vec method. It shows that the atomic feature vector proposed in this paper has a certain correlation between the features, and is a low-dimensional and densely distributed feature vector. Visual analysis and numerical experiments of material property prediction show that the atomic feature representation method based on deep learning extraction proposed in this paper can effectively characterize the material features and can be applied to the tasks of material band gap prediction.

Control strategy research of electric vehicle thermal management system based on MGA-SVR algorithm

The thermal management system is one of the important assemblies that ensure the secure operation of electric vehicles (EVs). Using intelligent algorithms to optimize the control strategy of the thermal management system can reduce energy consumption under the premise of effective heat dissipation of EVs. This paper attempts to construct the control strategy of EV thermal management system by coupling the modified genetic algorithm (MGA) and support vector regression (SVR). Firstly, the double-population adaptive mutation method and a novel optimization process are adopted to obtain MGA. Afterward, the performance of MGA is verified by four benchmark functions compared with three typical algorithms, which are genetic algorithm (GA), double-population genetic algorithm (DPGA), and quantum genetic algorithm (QGA). The results demonstrate that the accuracy and stability of MGA are obviously better than the other three algorithms. Secondly, MGA is applied to modify parameters of SVR kernel function, and the accuracy of MGA-SVR algorithm is verified by the Auto-MPG and Computer Hardware data sets. The mean square deviations of the SVR algorithm test set are 0.0186 and 0.0806, respectively, and the mean square deviations of the MGA-SVR algorithm test set are 0.0099 and 0.0054, respectively, which fully shows that MGA-SVR have more accurate forecasting capabilities. Finally, the thermal management system model of EV is built by the one-dimensional simulation software KULI. Under the Chinese working condition, fan speed which meets the cooling requirements of the motor and controller is obtained from the KULI model, and the database is constructed. Then, MGA-SVR is trained by database and employed to predict fan speed under the Chinese working condition and obtain control strategy of the thermal management system. Compared with traditional control strategy, the thermal management system based on MGA-SVR control strategy can not only meet the radiating requirements, but also effectively reduce the power consumption of fans.