Stock Price Prediction Using Long Short-Term Memory

Abstract

This document states the study of Long Short-Term Memory (LSTM) algorithms, for stock prediction system to predict the stock price effectively to minimize the difficulty of an individual in stock prediction. The main goal of this study is to use the LSTM algorithm with parameters to create a line graph that compares the real and forecasted stock prices. To train the algorithm, the project will employ the Python programming language in Jupyter Notebook and will use the Cross-Industry Standard Process for Data Mining (CRISP-DM) process methodology. There are two datasets will be used in the research which are Maxis and Maybank Stock price allocated from 2 November 2020 to 1 November 2021. To analyze the performance of the algorithms, the results would include a line graph and an indicator table such as Mean Square Error (MSE) and Mean Absolute Error (MAE). An individual or a firm can have a better understanding of prediction methods and how they affect the predictive result by conducting research. It found that LSTM model have high accuracy in the aspect of stock price prediction system and can predict the trend of the stock price in the future. However, LSTM model still cannot predict the stock price perfectly but just near the exact value. This is because there are many unpredictable and external factors will affect the stock price such as political issues and news. Therefore, In this research will suggest that act the LSTM model prediction as a assistance role or reference for an investor in decision making process instead of fully depend on it.