Accurate Prediction Stock Market Trends using BERT

This paper proposes an AI-based stock market trend prediction system using BERT to improve predictive accuracy compared to current models such as LIME. Materials and Two methods were compared Group 1, using LIME (Local Interpretable Model-agnostic Explanations) for interpretability-based prediction, and Group 2, using BERT (Bidirectional Encoder Representations from Transformers) for sophisticated contextual modeling. Both methods were tested on 20 samples using statistical significance at 80% power, 0.05 significance, and confidence level of 95%. The BERT-based model (Group 2) outperformed LIME (Group 1) with higher accuracy of 91.8% than LIME's accuracy of 85.7%, and statistically significant improvement (p < 0.05). The findings offer proves that BERT model provides higher accuracy, flexibility, and predictability and is thus a better option for stock market trend prediction.