Stock Trend Prediction

Abstract

The intent of this study is to use sophisticated methods of machine learning, particularly Long Short-Term Memory (LSTM) models, to improve the forecast precision for the value of the SENSEX index. It combines a large dataset with historical price statistics, technical signals (including SMA, EMA, OBV, and MACD), as well as sentiment evaluation (by employing the VADER sentiment analyzer) via news headlines. Pre-processing, normalization, and division of the data into training and testing sets are done. Metrics like Mean Absolute Percentage Error (MAPE) along with Mean Squared Error (MSE) are utilised in the training and assessment of different models built with LSTM. To maximise the efficiency of models, choosing attributes techniques like median range selection alongside K-Means clustering are implemented. The outcomes indicate how integrating different types of data and cutting-edge approaches to feature selection could end up innotable increases in forecasting precision.