Comparison of Support Vector Regression (SVR) and Long Short-Term Memory (LSTM) Methods for Meteorological Data Prediction in Nusa Tenggara
Andrew Castello Purba, Teny Handhayani, Janson Hendryli · 2025
This study compares Support Vector Regression (SVR) and Long Short-Term Memory (LSTM) for forecasting the meteorological data in Nusa Tenggara, on some variables i.e., temperature, rainfall, wind speed, and humidity. It used historical data from 1980 to 2023 from the Meteorology, Climatology, and Geophysics Agency in Indonesia. This research evaluates the performance of both methods based on accuracy and computational efficiency. The SVR demonstrated faster training times, making it suitable for applications requiring quick predictions. Meanwhile, the LSTM excels in capturing complex temporal patterns. SVR and LSTM achieved a maximum R2of 0.633 for forecasting average temperature, indicating similar performance. The findings provide insights into choosing the appropriate model based on specific use cases in meteorology.