Automatic lithology identification in meteorite impact craters using machine learning algorithms

Abstract

Abstract Identifying lithologies in meteorite impact craters is an important task to unlock processes that have shaped the evolution of planetary bodies. Traditional methods for lithology identification rely on time-consuming manual analysis, which is costly and limits the efficiency of rapid decision-making. This paper utilizes different machine learning algorithms namely Random Forest, Decision Tree, K Nearest Neighbors, and Logistic Regression with Grid Search to classify rock lithologies using data from the Bosumtwi impact crater in Ghana. A repeated stratified k-fold cross-validation method is applied to Grid Search to select the best combination of hyperparameters. The findings demonstrate that the Random Forest algorithm achieves the most promising results in classifying lithologies in the meteorite impact crater with an accuracy score of 86.89%, a recall score of 84.88%, a precision score of 87.21%, and an F1 score of 85.48%. The findings also suggest that more high-quality data has the potential to further increase the accuracy scores of the machine learning algorithm. In conclusion, this study demonstrates the significant potential of machine learning techniques to revolutionize lithology identification in meteorite impact craters, thus paving the way for their influential role in future space exploration endeavors.

Keywords

• Lithology classification
• Machine learning
• Random forest
• Bosumtwi impact crater
• Space exploration