Comparative Analysis and Implementation of Time Series Models for Air Quality Prediction in North Sumatra
Suvriadi Panggabean (a*), Faridawaty Marpaung (a), Zulfahmi Indra (a), Lasker P Sinaga (a)

a) Department of Mathematics, Faculty of Mathematics and Sciences, State University of Medan
*suvriadi[at]unimed.ac.id

Abstract

Air quality is a crucial issue that directly impacts public health and the environment, particularly in highly industrialized regions like North Sumatra. The ability to accurately predict air quality is crucial for early warning systems and policymaking. This study aims to conduct a comparative analysis of three time series forecasting models-Single Exponential Smoothing (SES), Double Exponential Smoothing (DES), and Autoregressive Integrated Moving Average (ARIMA)-to find the best model for predicting the Air Quality Index (AQI) in North Sumatra. The research methodology includes historical data collection, data preprocessing to handle missing values, implementation of the three models, and objective performance evaluation using test data. Model performance is measured using the Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE) metrics. The evaluation results show that the ARIMA model consistently provides the highest level of accuracy with the lowest RMSE and MAPE values compared to SES and DES. This indicates that air quality data in North Sumatra has complex temporal patterns that are more effectively captured by the ARIMA model. The selected ARIMA model is then used to generate short-term predictions, which can form the basis for developing a more responsive air quality monitoring system in the region.

Keywords: Air Quality- Time Series- Exponential Smoothing- ARIMA- North Sumatra

Topic: Mathematics and Computational System