Air Quality Classification in Urban Environment using Machine Learning Approach

Abstract

Air pollution comes from human activities that can threaten living things. It is affected by gasses including PM10, SO2, CO, O3, NO2 and others. Air pollution leads to dangerous diseases even death. Monitoring air quality is important task to understand pollution concentration. Air quality monitoring is better when it can classify whether air quality is habitable or not. This research proposes air quality classification using classification algorithms such as Logistic Regression, KKN, Decision Tree, and Random Forest algorithm. Dataset was taken from Jakarta’s open data for 12 months with several attributes including gas concentration and with several pre-processing steps. Based on experiment, decision tree model has the best accuracy to classify air quality level up to 100% with tuning several hyperparameters.

Urban Air Quality Classification Using Machine Learning Approach to Enhance Environmental Monitoring

“…Notably, there was a significant class imbalance observed. To address this, the "dangerous" and "very unhealthy" classes were merged into a single "unhealthy" class [27]. Consequently, the distribution of the final classes is as follows: "unhealthy" with 2,581 samples, "moderate" with 11,624 samples, and "good" with 2,910 samples.…”