LSTM MODELING WITH AN AUTOREGRESSIVE APPROACH FOR DAILY TEMPERATURE PREDICTION IN GRESIK REGENCY

Abstract

The zero-hunger program is one of the primary goals of the SDGs, especially in large countries like Indonesia, where hunger remains a serious issue. The agricultural sector plays a crucial role in addressing this problem. However, the effectiveness of this sector is highly dependent on climate changes, such as temperature. Therefore, this research aims to develop a daily temperature prediction model in Gresik Regency using the LSTM method with an autoregressive approach. This model is expected to assist farmers in optimizing planting and harvesting times. The autoregressive approach is applied by analyzing the ACF and PACF plots to determine the lags used as lookback parameters. The research results show that the LSTM model with five lookbacks and 150 epoch parameters provides the best outcomes, with an RMSE value of 0.50, MAE of 0.39, R2 of 0.69, and MAPE of 0.01.