Air Quality Classification Using the Backpropagation Neural Network Algorithm on DKI Jakarta Province ISPU Data
TL;DR
The effectiveness of the backpropagation neural network for classifying air quality based on DKI Jakarta ISPU data is demonstrated and provide a methodological basis for developing data-driven air quality monitoring and public health decision-support systems.
Abstract
Air quality is a critical environmental determinant of public health and quality of life. In Indonesia, air quality is commonly monitored using the Air Pollution Standard Index (ISPU), which classifies conditions according to pollutant concentrations. This study develops an air quality classification model using a backpropagation neural network and the DKI Jakarta ISPU dataset. The input variables comprised PM₁₀, PM₂.₅, SO₂, CO, O₃, and NO₂, while air quality was classified into three categories: Good, Moderate, and Unhealthy. The analytical procedure included data cleaning, min–max normalization, partitioning the dataset into 80% training and 20% testing subsets, and tuning the learning rate and number of hidden neurons. Model performance was evaluated using a confusion matrix, accuracy, Cohen’s kappa, sensitivity, and specificity. The optimal model, configured with a learning rate of 0.5 and six hidden neurons, achieved an accuracy of 97.67% and a kappa value of 0.933, indicating almost perfect agreement between predicted and observed classifications. High sensitivity and specificity across all categories further confirmed the model’s ability to distinguish air quality levels accurately. These findings demonstrate the effectiveness of the backpropagation neural network for classifying air quality based on DKI Jakarta ISPU data and provide a methodological basis for developing data-driven air quality monitoring and public health decision-support systems.