Development of stock price prediction system using Flask framework and LSTM algorithm

Abstract

Stock investment in Indonesia has been steadily growing in the past five years, offering profit potential alongside the risk of loss. Stockholders must analyze the stocks they intend to purchase. Stockholders often analyze stocks by observing patterns that occurred in the previous days to predict future prices. Therefore, a method is needed to simplify the process of analyzing the stock pattern. Although there are already several websites that have the concept of predicting stock prices, these websites do not utilize deep learning algorithms. This research aims to develop a stock price prediction website using deep learning algorithms, specifically the Long Short-Term Memory (LSTM) algorithm to help users predict stock prices. This research focuses on five banks with the highest market capitalization in Indonesia, namely Bank Central Asia, Bank Rakyat Indonesia, Bank Mandiri, Bank Negara Indonesia, and Bank Syariah Indonesia. The website utilizes Flask framework and LSTM. Flask is used to apply LSTM model to the website, while the LSTM can capture long-term dependencies in high-complexity data. The result of this research is a stock price prediction website application, where the prediction results are displayed through the website. The LSTM model for each stock has a Mean Absolute Percentage Error (MAPE) of less than 10%, which indicates that the model is “Highly accurate” based on the MAPE accuracy scale judgment.