Implementation of a Breakpoint Halfway Discretization to Predict Jakarta's Air Quality

Abstract

Despite the pandemic, Jakarta is one of the most polluted cities in the world. Knowing the daily air quality forecast aids the community, particularly Jakarta residents. Among these is the ability to protect oneself from dangerous air. The multinomial naive Bayes and the decision tree-ID3 methods are popular and perform well. Both of these strategies, however, require categorical variables. This need necessitates the implementation of a discretization technique for numerical variables. The purpose of this study is to predict Jakarta's air quality using the multinomial naive Bayes and decision tree method based on Particulate Matter 10 µg (PM10), Sulfur Dioxide (SO2), Nitrogen Dioxide (NO2), Ozone (O3), and Carbon Monoxide (CO). These continuous variables are discretized in two ways: using all midway breakpoints or halfway mixture breakpoints. The results indicated that the decision tree method with the mixture breakpoints halfway approach performed better than the multinomial nave Bayes method, with an accuracy of 98.90%, a specificity of 98.92%, a sensitivity of 75.00%, a precision of 75.00%, and an F1 score of 97.81%.