PEMANFAATAN PEMODELAN NEURAL PROPHET DALAM MEMPREDIKSI PARAMETER PM10 (STUDI KASUS DKI JAKARTA)

Abstract

Air quality is a serious issue in developing countries, especially in DKI Jakarta, driven by the increasing number of motor vehicles and industrial activities. This research examines the effectiveness of the Neural Prophet model in predicting PM10 concentrations as the primary air quality indicator. Daily data from DKI Jakarta's Air Quality Monitoring Station for the period 2018-2022 was used, with the model evaluated. Initial data underwent processing involving date format adjustments and the removal of irrelevant variables. The model was configured with predefined hyperparameters and tested using a holdout technique, splitting the data into a 95% training set and a 5% testing set. Model evaluation showed a significant reduction in errors, indicating effective learning. Time series data exhibited annual fluctuations, primarily peaking from June to October. The model's predictions aligned reasonably well with actual data, albeit with uncertainty at some points. This research demonstrates that Neural Prophet can predict PM10 concentrations with adequate accuracy, potentially serving as a tool for air quality management and planning in DKI Jakarta.