Enhanced Cascaded Greylag Goose Residual Deep Capsule Group Neural Networks-Based Prediction of Stock Market Prices from Social Sentimental Data

Stock market prediction using social sentiment data involves analyzing movements based on sentiments from social media, news articles, and other sources. This process extracts sentiment scores (positive, negative, neutral), aggregates them over time, and integrates them with historical stock prices. Traditional deep learning (DL) models face limitations like reduced accuracy, high error rates, and computational complexity. To address these, this work proposes an Enhanced Cascaded Greylag Goose Residual Deep Capsule Group Neural Networks (2C2GRDG2N) for accurate predictions. Input data from NIFTY-50 and real-time Twitter sentiment data are pre-processed, features are extracted and fused using Anisotropic Gaussian filtering, and optimal features are selected using Electric Eel Foraging Optimization Algorithm (EeFOA). The Cascaded Greylag Goose Residual Deep Capsule Group Neural Networks (2C2GRDG2N) then accurately predicts stock market trends. The Greylag Goose Optimization algorithm (GGOA) further improves prediction accuracy and reduces error rates, with the entire model implemented on the Python platform. The model attains 99.9% accuracy and 99.8% recall, outperforming existing methods. Implemented in Python, the model ensures cost-efficient and accurate predictions by leveraging optimal feature selection and advanced dimension reduction techniques.