Stock Market Prediction Using A 3-ML Fusion Approach

Abstract

Stock market prediction is still a significant challenge due to the non-linearity, dynamism, and volatility of the financial time series. Accurate stroke price prediction plays a major role in investment decisions, risk management, and financial planning. This research presents a novel 3-ML framework, which uses a combination of the Random Forest, Decision Tree and XGBoost Regression models to predict the next day's closing price of NIFTY-50 index. Moreover, a Random Forest (RF) classifier is used to classify market trends such as UP, DOWN or NEUTRAL, improving the decision-making process. Data from the market is preprocessed and enhanced by feature engineering through the use of technical indicators like moving averages and daily returns. The proposed framework implements a form of a Flask-based web application along with a MySQL database. It has been experimentally evaluated and shown that the framework developed has an accuracy of 97.7% and better stability of prediction and lower forecasting errors.