Implementasi ANN untuk Prediksi Angkutan Barang Kereta Api Berbasis Data Historis

Abstract

Railway transportation plays an important role in supporting goods distribution in Indonesia, particularly on the islands of Java and Sumatra. Changes in freight transportation volume over time form historical patterns that can be utilized for forecasting purposes. This study aims to implement an Artificial Neural Network (ANN) to predict railway freight transportation volume using historical data published by Statistics Indonesia (BPS) for the 2019–2026 period. The data were processed through preprocessing stages, including data cleaning, sliding window construction, Min-Max normalization, and splitting into training and testing datasets. The ANN model was developed using the Backpropagation algorithm and implemented in Google Colab utilizing TensorFlow and Keras libraries. The implementation results indicate that the model successfully learned historical data patterns, as demonstrated by the decreasing training loss during training and prediction results that closely followed the actual data trend. The findings indicate that ANN can be effectively applied as a time series forecasting approach for predicting railway freight transportation volume.