Gold Price Forecasting in Kuala Pilah, Negeri Sembilan, Malaysia Using Long Short-Term Memory (LSTM)

Abstract

Gold is the most popular investment in the world because it has proven to be the most effective haven in many countries. It is challenging to use technical analysis to predict gold's value. Many prediction problems involving time components require time series forecasting, an important topic in machine learning. This paperpresents a prototype for predicting the gold price in Kuala Pilah, Negeri Sembilan, Malaysia, using the Long Short-Term Memory (LSTM) time-series method. To address the problem, a dataset of daily gold prices was collected from Telegram Kedai Emas Nur Jannah and the Bullion Rates website. The main feature of the system is to predict the gold price and to visualise the predicted value. The waterfall method has been chosen as the project's methodology to ensure the project’s flow is correct. The predictive model was also evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). As a result, the system achieved an MAE of 0.108 at the daily time scale. The RMSE was 0.131 at the daily time scale, and the MAPE was 17%. The system can also improve the visualisation to make it more interactive and include another timescale, such as a daily timeframe.