Temporal Trend Analysis of Light Pollution Using Auto Regressive Integrated Moving Average (ARIMA) and Random Forest Models

Abstract Light pollution has become a global concern, affecting not only humans but also ecosystems, including animals and plants. Understanding its temporal trends and spatial distribution is crucial for mitigation efforts, especially in Asia, where urbanization is rapidly increasing. This research analyses light pollution trends using two models: Auto Regressive Integrated Moving Average (ARIMA) for time series forecasting and Random Forest for classification. The dataset consists of monthly light pollution data from several Asian countries, with key features including cloud cover, elevation, and magnitude. ARIMA was applied to predict future trends, while Random Forest was used to classify countries based on light pollution characteristics. Model performance was evaluated using Root Mean Squared Error (RMSE) and Mean Squared Error (MSE). The ARIMA model achieved an RMSE value of 1.58 indicates a relatively high accuracy in predicting future trends. The Random Forest model, after optimization, reached 99% classification accuracy with an MSE of 0.547, demonstrating its effectiveness in categorizing countries based on light pollution data. The results indicate that Eastern Asian nations, especially China and Japan, suffer from significant light pollution. In Southeast Asia, Indonesia exhibits the highest levels. ARIMA proves to be effective for long-term trend prediction, while Random Forest is adept at classification-based analysis. These insights are valuable for policymakers in environmental management and add to the broader research on light pollution.