New perspectives in the analysis of Stark width regularities and systematic trends

Abstract

Regularites and systematic trends among the sample of Stark widths obtained by using modified semiempirical method from the STARK-B database were analysed. Two different approaches are independently used -multiple regression method combined with simple cluster analysis, and random forest (RF) machine learning algorithm. Predicted values of Stark widths calculated with estimate formulae obtained from multiple regression method, and those values predicted by using RF algorithm, were compared with already known corresponding experimental Stark widths published elsewhere. Results of this analysis indicate that both of these methods can mostly predict new Stark width values within the acceptable range of accuracy.