Stock Price Prediction using Deep Learning Networks

Abstract

Accurate prediction of stock market prices plays a vital role in financial analysis and investment planning due to the highly dynamic and nonlinear nature of market behavior. This paper presents a comparative framework for stock price prediction using Artificial Neural Network (ANN) and Long Short-Term Memory (LSTM) models. Historical stock price datasets of multiple companies, including TATA, Tesla, Facebook, and Apple, are utilized to evaluate the predictive capability of both approaches. Prior to model training, the dataset undergoes preprocessing steps such as missing value removal, chronological sorting, Min–Max normalization, and division into training and testing subsets using an 80:20 ratio. The ANN and LSTM models are independently trained on the processed data and evaluated using Mean Squared Error (MSE) and prediction accuracy. Experimental analysis demonstrates that both models successfully capture stock price trends and generate predictions closely matching actual market values. However, the ANN model consistently produces lower MSE and higher prediction accuracy than the LSTM model across the evaluated datasets. Comparative graphical analysis further confirms the effectiveness of ANN in reducing prediction error while maintaining reliable forecasting performance. The proposed framework provides a practical and efficient solution for stock price prediction and can assist investors and financial analysts in making informed investment decisions through data-driven forecasting techniques.