Prediction of Antimicrobial Efficacy of Direct Cold Atmospheric Plasma Treatment via Ensemble Learning-based Regression Models

Cold atmospheric plasma (CAP) has found successful applications in diverse fields like biological decontamination, altering surface properties, and medical treatments. Its notable effectiveness against microorganisms underscores its established utility. Nevertheless, challenges persist, exemplified by the difficulty in comparing results from different research laboratories due to variations in CAP treatment conditions. In this context, the employment of machine learning (ML) algorithms presents a pivotal strategy for addressing these challenges. This study’s objective is to assess antimicrobial effectiveness using distinct parameters for direct CAP treatment, coupled with a range of ML algorithms. Constructing an original dataset from existing literature, we trained and evaluated this dataset using ensemble learning (EL) based regression models. Our analysis highlights the Bagged Tree approach as optimal, displaying a root mean square error (RMSE) of 1.291 and the coefficient of determination (R2) value of 0.72. Employing this approach facilitates swifter and more resource-efficient outcomes by curbing the time and resources consumed by laboratory experiments. Furthermore, it sheds light on the paramount parameters influencing microbial inactivation, thereby enhancing the efficiency of CAP applications. Employing ML techniques to analyze CAP’s antimicrobial effects under diverse conditions may obviate the necessity for redundant laboratory experiments, ensuring the acquisition of dependable outcomes.