Tectonic Setting Discrimination of Granites: Insights from Machine Learning

The training dataset provides a comprehensive and well-structured dataset comprising four distinct tectonic environments of granite. This dataset has been utilized to develop a machine learning (ML) classifier, specifically a random forest (RF) model, to distinguish among four granite tectonic settings: Continental Hotspot, Active Continental Margins, Intraplate Environment, and Continental Rift. For consistency across all datasets, these environments have been encoded with labels 0, 1, 2, and 3, respectively.

The prediction dataset contains trace element data of granites from various regions worldwide. It has been used to make predictions and assess the accuracy of the ML models by applying the previously trained RF model. Additionally, the dataset includes the formation ages of each sample.

The result dataset presents the RF model's classification outcomes for granite samples with previously unknown tectonic settings. The AGE1 results categorize the data into three geological periods—Archaean, Proterozoic, and Phanerozoic—based on the formation age of each tectonic environment. Meanwhile, the AGE2 results further subdivide the data into five geological eras: Archaean, Proterozoic, Paleozoic, Mesozoic, and Cenozoic, according to the predicted formation ages of the respective tectonic environments.