IMPLEMENTATION OF LONG SHORT-TERM MEMORY (LSTM) IN FORECASTING THE NUMBER OF TRAIN PASSENGERS IN JAVA ISLAND

Abstract

For certain Indonesians, trains are a particularly popular form of land transportation. Every year, during specific seasons, this mode of transportation consistently experiences a surge in passenger numbers. Due to this, it is necessary to make accurate predictions to make policies, such as whether additional carriages are needed. The selection of prediction methods will significantly impact policymaking. One of the methods currently being developed for prediction is related to machine learning. This study aims to implement a forecasting method using machine learning that can be used to predict time series data. The machine learning used in this study is the Long Short-Term Memory (LSTM) method. In this study, we used time series data on the number of train passengers. The data used is secondary data from the Statistics Indonesia (BPS). The data analysis process in this study uses Python software. The results of this analysis show that the LSTM model has a high level of accuracy in prediction, indicated by the mean squared error value of 2,941,137.156 and MAPE of 0.07%. Forecasts show a gradual increase in the number of passengers, starting from 32,381 people in the first month to 33,068 people in the third month. These results indicate that the LSTM model is thought to be effective in predicting changes in the number of train passengers, and further research is needed to verify this assumption.