Comparison of Stock Mutual Fund Price Forecasting Results Using ARIMA and Neural Network Autoregressive Model

Abstract

Stock mutual funds gained popularity among the public as an investment alternative due to the convenience they offer, especially for beginner investors who have limited time and investment knowledge. Compared to money market and bond mutual funds, these mutual funds offer higher potential returns but also come with higher risks due to value fluctuations, so forecasting stock mutual fund prices is essential to minimize losses. Since stock mutual fund prices is time series data, this research employs two forecasting models such as Autoregressive Integrated Moving Average (ARIMA) and Neural Network Autoregressive (NNAR). The objective of this research is to determine the best-performing model between ARIMA and NNAR, and compare their forecasting accuracy using the Mean Absolute Percentage Error (MAPE). The data used consists of daily closing prices of stock mutual funds from March 1, 2022, to March 31, 2025, with the criteria that the selected issuers have been operating for more than five years. The results of this research show that the best ARIMA and NNAR for the RNCN are ARIMA([1],1,0) and NNAR(2,2); for TRAM are ARIMA(0,1,[1]) and NNAR(4,1); for SCHRP are ARIMA(0,1,[1]) and NNAR(4,2); for MICB are ARIMA([1],1,0) and NNAR(2,2); and for BNPP are ARIMA([1],1,0) and NNAR(5,1). The MAPE values in the same order are 6.83% and 5.49%; 6.53% and 5.75%; 8.57% and 7.10%; 8.39% and 8.75%; 8.51% and 7.30%. Based on the comparison, NNAR outperformed ARIMA in four out of five mutual funds, with lower MAPE values and also marked by the ARIMA model tend to produce stable or unchanging values over the long term. The results of this research are expected to assist investors in consederating by choosing NNAR model, both in the short and long term, to obtain better stock mutual fund price forecasts.