Тип публикации: статья из журнала
Год издания: 2024
Идентификатор DOI: 10.32014/2024.2518-170X.464
Ключевые слова: rock classification, machine learning in geology, analysis of oxides in rocks, classification accuracy, model evaluation metrics, importance of features
Аннотация: Automation of rock classification by chemical composition is an important task in geology, ecology, mining and construction. Traditional methods based on visual characteristics and field studies are not effective enough to process large amounts of data. The paper describes the development of a model for classifying rocks based on tПоказать полностьюheir chemical composition using logistic regression. The dataset used was a dataset of key oxide concentrations of different rock types. The quality of the model was assessed using the accuracy, recall, precision and F1- score metrics, as well as visualization of the results using confusion matrices and heat maps of rock distribution by geographic zones. The model effectively used chemical features to identify these rocks. However, for rare or similar rocks (diorite, granodiorite, rhyolite), the accuracy was lower, indicating the need for further work to improve the model. Heat map of rock distribution by region (Washington, Oregon) demonstrates the correlation between rock type and geological features (volcanic activity, subduction). The model has proven effective in classifying common rock types. The results can be applied in practice to speed up and improve the accuracy of geological studies, especially when processing large amounts of data, for example, to automate the analysis of drilling results and determine rock types over large areas. The obtained data on the distribution of rocks in a region can help in understanding geological history and predicting possible mineral deposits.
Журнал: Известия Национальной академии наук Республики Казахстан. Геология Казахстана
Выпуск журнала: Т. 6, № 468
Номера страниц: 114-125