PEMODELAN DERET WAKTU NILAI ISPU PM2,5 DI SURABAYA, MALANG, DAN SIDOARJO MENGGUNAKAN METODE MLR, MARS, DAN EXPONENTIAL SMOOTHING

Abstract

Air pollution, particularly fine particulate matter (PM2,5), remains an environmental issue affecting public health, especially in areas with high anthropogenic activity such as Surabaya City, Malang City, and Sidoarjo Regency. This study aims to determine the time series modelling method with the best accuracy in predicting the Air Pollutant Standard Index (ISPU) value of PM2,5, to examine the influence of meteorological factors (rainfall, temperature, humidity, wind speed, and wind direction) on ISPU PM2,5 values, and to predict ISPU PM2,5 values for the next one-year period in the three study areas. The data used were secondary data on ISPU PM2,5 values obtained from the East Java Provincial Environmental Agency and meteorological data from BMKG for the period of January 2023-December 2025, analyzed using Multiple Linear Regression (MLR), Multivariate Adaptive Regression Splines (MARS), and Exponential Smoothing methods with the assistance of R software. The results showed that the best performing model differed across regions: MLR was identified as the best model for Surabaya City and Malang City, while MARS was the best model for Sidoarjo Regency, based on the lowest MAE, MAPE, AIC, and Adjusted R-Squared values. Feature importance analysis revealed that average temperature was the most influential meteorological factor on ISPU PM2,5 values in Surabaya and Malang, whereas wind direction and rainfall were the most dominant factors in Sidoarjo. Prediction results for 2026 indicate that ISPU PM2,5 values in the three regions tend to remain in the moderate category, with Sidoarjo Regency showing the potential to approach the unhealthy for sensitive groups category. The findings of this study are expected to serve as a consideration for local governments in formulating air pollution control strategies.