CLASSIFICATION OF EXOPLANETS BASED ON A MACHINE LEARNING MODEL
Authors:
1.
Kazan National Research Technical University named after A.N. Tupolev
2.
Kazan National Research Technical University named after A.N. Tupolev
3.
Kazan National Research Technical University named after A.N. Tupolev
Type:
Article
Pages:
from 137
to 143
Status:
Published
Received:
03.12.2025
Accepted:
08.04.2026
Published:
05.05.2026
Language:
Russian
Keywords:
EXOPLANETS, MACHINE LEARNING, CLASSIFICATION, MACHINE LEARNING MODEL, RANDOM FOREST
Abstract:
This article presents a study of the feasibility of using machine learning methods to classify exoplanets based on astronomical data. The object of the study is exoplanets, and the subject is approaches to constructing and interpreting exoplanet classification models based on their physical and orbital characteristics. The Exoplanet Classification Dataset contains 19,761 observations and 16 features, including stellar parameters, photometry data, and the resulting class label. The dataset is divided into samples: a training sample (12,646 objects - 64 %), a validation sample (3,162 objects - 16 %), and a test sample (3,953 objects - 20 %). The number of objects is distributed among classes: class 0 - 6,311 objects, class 1 - 7,413 objects, class 2 - 6,015 objects, and class 3 - 22 objects. Data preprocessing was performed, including feature normalization, gap handling, and class balancing using the Synthetic Minority Oversampling Technique (SMOTE). The Random Forest algorithm was selected to implement the machine learning model. A comparison of the Random Forest algorithm with other classification algorithms is described: logistic regression, support vector machine (SVM), gradient boosting, and a simple neural network (MLP). A comparative analysis of the SMOTE method is conducted. The adequacy of the developed model was assessed using the precision and recall metrics. The final accuracy of exoplanet classification on the test set was 75 %. Based on the resulting models, the importance of physical features influencing the classification of exoplanets into different types was determined, which allows the results to be interpreted not only from a machine learning perspective, but also from an astrophysical perspective. The developed machine learning model forms the basis for intelligent systems supporting scientific discoveries in modern space exploration.
This article presents a study of the feasibility of using machine learning methods to classify exoplanets based on astronomical data. The object of the study is exoplanets, and the subject is approaches to constructing and interpreting exoplanet classification models based on their physical and orbital characteristics. The Exoplanet Classification Dataset contains 19,761 observations and 16 features, including stellar parameters, photometry data, and the resulting class label. The dataset is divided into samples: a training sample (12,646 objects - 64 %), a validation sample (3,162 objects - 16 %), and a test sample (3,953 objects - 20 %). The number of objects is distributed among classes: class 0 - 6,311 objects, class 1 - 7,413 objects, class 2 - 6,015 objects, and class 3 - 22 objects. Data preprocessing was performed, including feature normalization, gap handling, and class balancing using the Synthetic Minority Oversampling Technique (SMOTE). The Random Forest algorithm was selected to implement the machine learning model. A comparison of the Random Forest algorithm with other classification algorithms is described: logistic regression, support vector machine (SVM), gradient boosting, and a simple neural network (MLP). A comparative analysis of the SMOTE method is conducted. The adequacy of the developed model was assessed using the precision and recall metrics. The final accuracy of exoplanet classification on the test set was 75 %. Based on the resulting models, the importance of physical features influencing the classification of exoplanets into different types was determined, which allows the results to be interpreted not only from a machine learning perspective, but also from an astrophysical perspective. The developed machine learning model forms the basis for intelligent systems supporting scientific discoveries in modern space exploration.
Keywords:
EXOPLANETS, MACHINE LEARNING, CLASSIFICATION, MACHINE LEARNING MODEL, RANDOM FOREST
EXOPLANETS, MACHINE LEARNING, CLASSIFICATION, MACHINE LEARNING MODEL, RANDOM FOREST
