A MULTI-LAYER PERCEPTRON NEURAL NETWORK FOR PREDICTING MOBILE BANKING TRANSACTION

Abstract

Mobile banking has become increasingly prevalent in today's digital age, providing convenient and accessible financial transactions for users. This research used a multi-layer perceptron (MLP) of neural network model for forecasting mobile banking transaction volumes in Indonesia. Utilizing secondary data sourced from Bank Indonesia's official website, a dataset comprising 133 time series observations from January 2013 to January 2024 was analyzed. The data was divided into training and testing sets, with proportions of 80% and 20%, respectively. Analysis was conducted using RStudio. The MLP was configured with 5 hidden nodes and 20 repetitions. The evaluation of results revealed noteworthy differences in model performance between the training and test sets. The MLP model delivered high forecasting accuracy with minimal error, evidenced by a Mean Absolute Error (MAE) of 754.294, a Mean Absolute Percentage Error (MAPE) of 1.163, and a Mean Absolute Scaled Error (MASE) of 0.027. These results confirm the MLP model's reliability as an effective tool for time series forecasting. Forecasting results indicate a significant upward trend in mobile banking transaction volumes over the next 24 months (February 2024–January 2026). The highest volume is projected in December 2025, reaching 2,065,309 transactions, with minor fluctuations observed in some months. These findings provide valuable insights for financial institutions to strategize and allocate resources effectively to accommodate the anticipated growth in mobile banking