Comprehensive Analysis of Mudflow Data Using Machine Learning Methods

The paper presents a comprehensive analysis of mudflow basin parameters, conducted using machine learning methods. For the northern slope of the Greater Caucasus, data on the main parameters of mudflow basins were analyzed to build models that allow forecasting mudflows with certain characteristics. A set of machine learning methods was used (clustering, search for association rules, logistic regression, etc.). Key factors of mudflows were identified, models were developed for classifying mudflow types and predicting the volume of one-time removal of material, and a number of association rules with high reliability were identified that describe the relationships between factors influencing mudflow processes. The obtained results show great potential in the application of learning in the tasks of analysis and forecasting of mudflow processes. Ultimately, this will allow, based on the addition of mudflow data and updating of existing mudflow maps, to develop more effective measures to reduce the impact of mudflows on the environment to a minimum.