Perbandingan Prediksi Emisi Gas Buang Kendaraan Bermotor menggunakan Regresi Linear dan Long-Short Term Memory

Abstract

Vehicle exhaust emissions are a major source of air pollution in urban areas. This study compares two predictive approaches, Linear Regression and Long Short-Term Memory (LSTM), to model vehicle exhaust gas emissions such as CO, NO₂, NH₃, HC, and PM using data collected over 7 days. The data includes environmental parameters (temperature and humidity) as well as vehicle parameters (speed and distance traveled). The models were trained using the first 7 days of data and tested with limited test data. The evaluation results show that Linear Regression performs very well, particularly for gases with stable patterns like CO and PM, with R² values above 0.98 and low prediction errors. In contrast, LSTM is more effective in predicting gases with dynamic fluctuations, such as NH₃ and HC, although slightly lagging due to the limited historical data. This study concludes that Linear Regression is more suitable for gases with stable patterns and limited data, while LSTM excels in long-term predictions with more complex data. The results of this study can serve as a reference for developing sensor-based emission prediction systems to support adaptive pollution control policies and carbon trading.