Forecasting Data Time Series Menggunakan MLP dan LSTM untuk Memprediksi Jumlah Produksi Bir

Rachmatullah, Muhammad Ibnu Choldun

TIN: TERAPAN INFORMATIKA NUSANTARA
Vol 6 No 4 (2025): September 2025

Time series data forecasting is an important approach in various sectors such as finance, energy, and healthcare. As technology advances, deep learning methods such as Multi-Layer Perceptron (MLP) and Long Short-Term Memory (LSTM) are increasingly being used to improve prediction accuracy. This study compares the performance of these two methods in forecasting a time series dataset of monthly beer production in Australia. The model was trained and tested using a 70% training and 30% testing data split. Performance evaluation was based on the Root Mean Square Error (RMSE) value after 10 experimental repetitions. The results show that MLP has a lower RMSE value and a smaller standard deviation than LSTM, both on the training and testing data. This indicates that MLP is more stable and efficient in handling datasets with simple patterns and low complexity, while LSTM tends to require more intensive tuning and has a higher risk of overfitting. Therefore, MLP is recommended as a lighter and more consistent alternative forecasting method for similar data scenarios.