Mutia Delina, Alpon Sepriando, Rizqi Chiesa Nurdiansyah
Short-term weather nowcasting plays a critical role in various sectors, including agriculture, transportation, and disaster mitigation. This study aims to implement a machine learning approach for short-term weather prediction and to identify the most influential weather parameters affecting prediction accuracy. A quantitative experimental method was employed: the Random Forest algorithm with the RandomizedSearchCV technique. The dataset consisted of historical weather observations from the Soekarno-Hatta Meteorological Station, from January 1 to December 31, 2024. The results show that the model achieved F1-scores above 0.94 for all weather classes during training and validation. However, the imbalanced dataset test that reflects actual weather, the model reached an F1-score of 0.967 for predicting no-rain conditions one hour ahead, while performance for rain classes was significantly lower, ranging from 0.21 to 0.41. Current weather conditions emerged as the most influential feature, with F-scores of 3508.58 (1 hour), 929.09 (2 hours), and 397.43 (3 hours), followed by rainfall (515.34), humidity (173.98), and temperature (56.74). This study successfully demonstrates the application of machine learning for short-term weather nowcasting, providing reliable predictions under general weather conditions. © Published under licence by IOP Publishing Ltd.
Physics Department, Faculty of Mathematics and Science, Universitas Negeri Jakarta, Jl. Rawamangun Muka, Jakarta Timur, 13220, Indonesia; The Remote Sensing Imagery Management, Meteorological Climatological and Geophysical Agency, Jl. Angkasa, Jakarta Pusat, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock