Real-Time Mechanism Based on Deep Learning Approaches for Analyzing the Impact of Future Timestep Forecasts on Actual Air Quality Index of PM10

• Mapping seasonal patterns in AQI PM 10 concentration in air quality in Jakarta. • Integrating AQI PM 10 analysis into a standard air pollution index for Jakarta. • Developing a real-time deep learning-based air quality prediction system. • Combining ARIMA and GRU for handling missing data and accurate forecasting. • Analyzing the impact of forecast timesteps on AQI PM 10 prediction. The air quality in Jakarta, particularly concerning PM 10 particulate matter, has become a serious concern due to its significant impact on health and the environment. The increase in pollution in this city is often triggered by industrial activities and seasonal factors, such as forest fires, which create challenges in air quality management. This study aims to map seasonal patterns in AQI PM 10 concentrations and to develop a DL-based air quality prediction system that can operate in real-time, while also considering the impact of future forecasting timesteps on actual values. Additionally, this goal supports effective mitigation efforts and provides new insights for decision-making in addressing urban pollution. We propose an approach using GRU for time series forecasting, combined with an ARIMA model for imputing missing data. Utilizing air quality index data from Jakarta over more than 13 years, this research identifies consistent seasonal patterns, with PM 10 concentrations peaking between May and October. Most days in this dataset fall into the “Moderate” category of the AQI, although there is considerable variation influenced by seasonal phenomena and industrial activities. Results show that the GRU model effectively captures fundamental patterns in the AQI PM 10 data, although predictive accuracy tends to decrease as the forecasting interval increases. This study also provides new insights into the application of the GRU model for multi-timestep forecasting, focusing on short- and medium-term predictions that can be used as a decision-making tool for urban pollution mitigation.