Peramalan Deret Waktu Penjualan Bulanan Menggunakan Pendekatan Deep Learning Metode Long Short-Term Memory

Permatasari, Intan
Puspitodjati, Sulistyo

TIN: TERAPAN INFORMATIKA NUSANTARA
Vol 6 No 8 (2026): January 2026

Companies often perform sales forecasting to determine inventory levels in order to control stock and ensure product availability prior to sales activities. This study aims to compare the performance of the Autoregressive Integrated Moving Average (ARIMA) method and the Long Short-Term Memory (LSTM) method in forecasting monthly sales quantities. This research employs a quantitative experimental approach, where the LSTM model as part of a deep learning framework is used to learn structured time series patterns. The dataset consists of historical monthly demand data from 2012 to 2018 using a one time step prediction approach. Model performance is evaluated using Root Mean Square Error (RMSE) as the accuracy metric. The results show that the LSTM model achieves an average RMSE of 5.39 across ten experimental runs, which is lower than that of the ARIMA method reported in previous studies using the same dataset. These findings indicate that the LSTM method provides better forecasting performance and can serve as a reliable basis for monthly inventory planning decisions.