Enhancing rainfall prediction in SVM by incorporating the autocorrelation function (ACF)

• Published in: AIP Conference Proceedings
• Main Author: Marsani M.F.; Someetheram V.; Mohd Kasihmuddin M.S.; Mohd Jamaludin S.Z.; Mansor M.A.; Badyalina B.; Jan N.A.M.
• Format: Conference paper
• Language: English
• Published: American Institute of Physics 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85203123903&doi=10.1063%2f5.0223837&partnerID=40&md5=3e177f1ad0d084b3f1fee7059b9875a2

Accurate prediction of rainfall is crucial for anticipating and preparing for potential disasters such as floods or droughts in specific areas. This study introduces the Autocorrelation Function (ACF) method as a potent tool for modeling these dependencies within Pahang rainfall data. By examining the temporal relationships between current and past values, the ACF method yields valuable insights. This study leverages ACF to forecast rainfall at specific time intervals, capturing crucial temporal patterns within the daily, weekly maximum and monthly maximum interval. Our findings highlight the remarkable accuracy of the ACF method in enhancing Support Vector Machine (SVM) model performance, especially when there is a strong relationship between the current and past values of the rainfall series. Moreover, the forecasting analysis reveals that incorporating the ACF method into the SVM model surpasses conventional techniques, as evidenced by the lowest root mean square error (RMSE) values. In conclusion, this study advances our comprehension of rainfall dynamics in Pahang, strengthening the capabilities for improved forecasting and risk mitigation strategies. © 2024 Author(s).