Prediksi Jumlah Kejadian Banjir Bulanan di Indonesia Berdasarkan Analisis Long Short Term Memory

Alfredi Yoani (Universitas Airlangga, Indonesia)
Sediono Sediono (Universitas Airlangga, Indonesia)
M. Fariz Fadillah Mardianto (Universitas Airlangga, Indonesia)
Elly Pusporani (Universitas Airlangga, Indonesia)

G-Tech : Jurnal Teknologi Terapan
Vol 7 No 4 (2023): G-Tech, Vol. 7 No. 4 Oktober 2023

Floods are among the most common and dangerous natural disasters worldwide, leading to loss of life and economic instability. In Indonesia, floods have been the most frequently occurring natural disaster since 2009. The high frequency underscores the urgency of predicting the number of natural disaster events to assist the government and the public in taking appropriate mitigation measures, as well as contributing to the achievement of Sustainable Development Goal 15 regarding Terrestrial Ecosystems. The method used to predict the monthly occurrence of floods in Indonesia is Long Short Term Memory (LSTM). LSTM was chosen for its ability to process sequential data over a long period of time. Upon analysis, highly accurate forecasting results were obtained, with a Mean Absolute Percentage Error (MAPE) of 8.04%, a Root Mean Square Error (RMSE) of 5.991. The model is also proficient at estimating training data, with an value of 95.71%.