Analysis Of Closing Price Forecasting For Bank Negara Indonesia Stocks Using A Hybrid K-Means Backpropagation Neural Network Model

Abstract

Stock price movements are dynamic in nature, and their active patterns make them difficult to predict accurately. These conditions have driven the development of forecasting models capable of recognizing data characteristics before generating predictions with a low error rate. This study aims to analyze how the hybrid K-Means and Backpropagation Neural Network model works in forecasting stock closing prices. The dataset used consists of historical closing prices of Bank Negara Indonesia (BNI) stock over a 5-year period, comprising a total of 1,447 data points. Based on the results of this study, the model achieved an RMSE of 109.26, an MAE of 83.30, and a MAPE of 2.1458%, indicating that the model is capable of predicting stock prices with a low error rate based on these evaluation metrics. The addition of klasters to the neural network training process helps the model recognize data characteristics, resulting in more stable predictions. This study provides a hybrid-model-based approach to stock closing price forecasting that can serve as a reference for developing time-series prediction models in the field of finance.