Prediksi Intensitas Curah Hujan di Kota Surabaya Menggunakan Principal Component Analysis PCA Dan Long ShortTerm Memory LSTM

Abstract

Surabaya frequently experiences high rainfall, which has the potential to cause flooding due to its suboptimal drainage system. Therefore, accurate rainfall prediction is crucial for disaster mitigation and urban planning. This study aims to measure the extent to which Principal Component Analysis (PCA) influences the accuracy improvement of the Long Short-Term Memory (LSTM) model in predicting rainfall intensity in Surabaya. PCA is used to extract features from meteorological variables such as air pressure, temperature, humidity, sunlight duration, and wind direction to reduce data dimensionality and enhance model efficiency. This study evaluates three scenarios with different numbers of PCA variables: 5, 6, and 7 variables. The results indicate that the scenario with 6 PCA variables achieves the best performance, with an MAE of 0.012, RMSE of 0.033, dan MSE of 0.0010. This scenario demonstrates an optimal balance between the number of variables and model accuracy. The findings confirm that dimensionality reduction using PCA can improve the efficiency and accuracy of LSTM-based rainfall prediction models; however, the number of selected variables must be optimal to avoid significant information loss. The developed model is expected to serve as a reference and guideline for improving rainfall prediction accuracy.