International Journal of Scientific Research and Engineering Development

Share market forecasting remains a difficult computational problem because price series are noisy, non-stationary, and highly sensitive to short-term shocks. This paper presents an AI-based academic prototype that converts historical open, high, low, close, and volume records into structured forecasts and decision-support signals through a disciplined time-series pipeline. The study compares three model families—Linear Regression, Random Forest, and Long Short-Term Memory networks—under a common workflow consisting of chronological data collection, cleaning, technical indicator engineering, scaling, time-aware train-validation-test splitting, and comparative evaluation. In addition to raw market fields, the prototype uses engineered features such as returns, moving averages, relative strength index, moving average convergence divergence, volatility measures, and lag variables. The discussion shows that Linear Regression is valuable as a transparent baseline, Random Forest is effective for nonlinear interactions in feature-rich tabular inputs, and LSTM is particularly suited to sequential dependency learning. The paper emphasizes practical safeguards against leakage, overfitting, and concept drift, and it interprets forecasting as a tool for academic analysis and decision support rather than a guarantee of profit. The resulting framework is reproducible, extensible, and appropriate for student-level experimentation in financial analytics.

Vansh Vashist, Mayur Narang, Mr. Karmbir,"Integrated Computational System for Share Market Forecasting: Comparative Evaluation of Linear Regression, Random Forest, and Long Short-Term Memory Architectures" International Journal of Scientific Research and Engineering Development, V9(3): Page(546-549) May-June 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.