IMPLEMENTASI LONG SHORT-TERM MEMORY UNTUK PREDIKSI HARGA LPG 3 KG

Abstract

Fluctuations in the price of 3 kg LPG in East Kalimantan affect economic stability, especially for households and small businesses. This uncertainty complicates planning and decision-making, thus requiring an accurate predictive system. This study aims to apply the Long Short-Term Memory (LSTM) method to predict LPG prices using time-series data. The dataset consists of 1,853 daily records from 2018 to 2025. Preprocessing includes handling missing values with interpolation and forward filling, normalization with Min-Max Scaling, and data transformation with a sliding-window approach. The LSTM model consists of two hidden layers and is optimized using the Adam optimizer. Model evaluation is conducted using Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The results show an MAE of 994.14 and an RMSE of 2211.17, indicating that the model's predictions are close to the actual values. Therefore, LSTM is an effective approach for LPG price prediction, supporting better decision-making