COMPOSITION OF MACHINE LEARNING METHODS IN THE ANALYSIS OF THERMALLY STIMULATED LUMINESCENCE KINETICS

Relevance Thermally stimulated luminescence (TSL) kinetics analysis is important for studying materials with special optical properties used in sensors and imaging systems. Traditional TSL analysis methods are often ineffective when dealing with complex data. Combining clustering and regression algorithms increases the accuracy of forecasts and helps to identify hidden patterns. Aim of research Aim of this work is to create a s􀀀stem for predicting the kinetics of polyarylene phthalides using machine learning methods. To achieve this goal, it is planned to develop a methodology for processing experimental data, including clustering and composition of regression models. Research methods To process the data and identify patterns, the KMeans clustering method was used, which made it possible to divide the experimental data into groups with different behavioral features. Various regression methods, including linear regression and ensemble approaches such as XGBRFRegressor, were used to model the dependence of luminescence intensity on temperature and time parameters. To determine the optimal number of clusters, approaches such as the elbow method, silhouette analysis, and the Davies-Bouldin method were used, which provided reliable data partitioning. Results A system for predicting the kinetics of polyarylene phthalides based on machine learning has been developed. The approach divided the data into groups with different characteristics, making it possible to create individual regression models for each cluster and improve the accuracy of the forecast.