Automatic lithology identification in meteorite impact craters using machine learning algorithms

Area of Science:

• Planetary Science
• Geology
• Computer Science

Background:

• Lithology identification in impact craters is crucial for understanding planetary evolution.
• Traditional manual methods are slow, costly, and inefficient for rapid analysis.
• Machine learning offers a potential solution to automate and improve lithology classification.

Purpose of the Study:

• To evaluate machine learning algorithms for classifying rock lithologies in the Bosumtwi impact crater.
• To compare the performance of Random Forest, Decision Tree, K Nearest Neighbors, and Logistic Regression algorithms.
• To identify the most effective machine learning model for this task.

Main Methods:

• Utilized data from the Bosumtwi impact crater in Ghana.
• Applied Random Forest, Decision Tree, K Nearest Neighbors, and Logistic Regression algorithms.
• Employed Grid Search with repeated stratified k-fold cross-validation for hyperparameter tuning.

Main Results:

• The Random Forest algorithm achieved the highest accuracy (86.89%), recall (84.88%), precision (87.21%), and F1 score (85.48%).
• The study indicates that higher quality data could further improve machine learning model performance.
• Machine learning demonstrates significant potential for efficient and accurate lithology identification.

Conclusions:

• Machine learning techniques, particularly Random Forest, show great promise for revolutionizing lithology identification in impact craters.
• This automated approach can significantly improve the efficiency and accuracy of geological analysis for planetary bodies.
• The findings support the integration of machine learning in future space exploration missions for rapid data analysis.