Comparative modelling for level of rainfall using data mining techniques

• Published in: AIP Conference Proceedings
• Main Author: Mokhranim N.S.; Derasit Z.; Mansor M.M.; Osman B.M.; Kamarudin N.
• Format: Conference paper
• Language: English
• Published: American Institute of Physics 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85203154941&doi=10.1063%2f5.0224239&partnerID=40&md5=b69e54124fcf3f111e07b57302db32f7

The effects of climate change are becoming increasingly noticeable in Malaysia which has been experiencing extreme weather events in recent years. The phenomenon of extreme weather has resulted in unpredictable variations in rainfall distribution, highlighting the critical need for accurate rainfall prediction. In addition, accurate rainfall prediction in airport meteorological stations is equally important because airports are considered one of the top priorities in meteorological stations. Therefore, the aim of this study is to determine the best rainfall level prediction model, focusing on three airport meteorological stations; Subang, Kota Bharu, and Senai, by using 15 years (2007-2021) of rainfall data. Two data-mining models, Support Vector Machine and Artificial Neural Network are employed to build the prediction model with different approaches, including data splitting mechanism and feature selection methods (optimize selection, forward selection, and backward elimination). The performance of the best model is assessed by using misclassification rate, accuracy, RMSE, and MAE. The results indicate that the Support Vector Machine, combined with optimized feature selection using a 90:10 ratio, outperforms the Artificial Neural Network models in terms of accuracy (76.36%), misclassification rate (23.64%), RMSE (48.60%) and MAE (23.60%) values. This developed model could serve as a reliable tool for rainfall prediction, benefiting agencies responsible for air traffic control in their planning and decision-making processes. © 2024 Author(s).