Application Of Principal Component Analysis (PCA) Algorithm And Long Short-Term Memory (LSTM) Network For Dasarian Rainfall Forecasting In Pekanbaru Region

Pekanbaru City, as the center of government, education, trade, and industrial activities, faces challenges related to tropical climate change, including rainfall fluctuations that affect disasters such as forest fires, floods, and landslides. The Meteorology, Climatology, and Geophysics Agency (BMKG) have noted an increase in rainfall that could trigger hydrometeorological disasters in the region. This research aims to address this problem by proposing a dasarian rainfall prediction model using a deep learning approach. The Long Short-Term Memory (LSTM) method was chosen for its ability to capture temporal patterns in time series data. At the same time, Principal Component Analysis (PCA) was used to reduce the dimensionality of the data and identify the most relevant features, such as temperature, humidity, and wind speed. The historical climate data of Pekanbaru from 2018 to 2023 in time series form was used as the model input. Results show that the PCA-LSTM model performs better than a single LSTM, with error values of RMSE 5.97, MAE 3.74, MAPE 0.10%, and R-squared 0.94. This PCA-LSTM combination can explain 87% of the cumulative variance, which contributes significantly to explaining rainfall variability and capturing the seasonal pattern of rainfall, so it can help mitigate hydrological disasters in Pekanbaru.