STOCKIQ: Comparative Analysis of Data Driven Models for Historical Stock Market Prediction

Abstract: Stock market prediction is a challenging task due to its dynamic and nonlinear nature.This project presents StockIQ, a machine learning-based system for predicting next-day stock prices.The system utilizes historical stock data along with multiple technical indicators for feature engineering.Models such as Random Forest, Linear Regression, XGBoost, and LSTM are used for prediction.A time-series approach is followed to ensure proper training and testing without data leakage.The system also includes a real-time analysis module and a historical backtesting module.Performance is evaluated using metrics like RMSE, MAE, MAPE, and directional accuracy.Results show that ensemble models provide better prediction accuracy compared to basic models.The system demonstrates practical applicability for stock analysis and forecasting.However, predictions are limited by market volatility and external influencing factors.

Keywords: Stock Price Prediction, Machine Learning, LSTM, XGBoost, Technical Indicators, Time Series Analysis, Backtesting