Prediction of asthma outpatients using cumulative particulate matter and machine learning algorithms: a case study in Adiyaman, Turkey

Abstract

Following the 2023 earthquake in Adıyaman, Turkey, particulate matter (PM10) levels saw a significant rise, prompting the need to develop a model linking these levels to the number of asthma cases in the city. Using PM10 data from the Adıyaman urban region, we built machine learning models to estimate asthma prevalence. The k-nearest neighbour, random forest, and linear regression algorithms were employed for this purpose, with random forest outperforming the others, achieving an R-value of 0.92. The k-nearest neighbour and linear regression techniques followed with R-values of 0.81 and 0.64, respectively. The model's performance was further enhanced by incorporating cumulative PM10 pollution data as input parameters. This study is notable as it is only the second in the literature to estimate the asthma prevalence using air pollution data, and it achieved a higher accuracy rate than the previous study.