ENHANCING MARKET TREND FORECASTING WITH EXPLAINABLE AI: A COMPARATIVE ANALYSIS OF DEEP LEARNING MODELS AND INTERPRETABILITY TECHNIQUES

Lokendra Singh Sisodia; Ankur Khare

doi:10.29121/shodhkosh.v5.i3.2024.5185

Authors

Lokendra Singh Sisodia Research Scholar,Computer Science & Information Technology Department, Rabindranath Tagore University, Raisen, (M.P.), India
Dr. Ankur Khare Assistant Professor,Computer Science & Information Technology Department, Rabindranath Tagore University, Raisen, (M.P.), India

DOI:

https://doi.org/10.29121/shodhkosh.v5.i3.2024.5185

Abstract [English]

In financial analytics, investing and managing risk is deeply connected with forecasting market trends which is one of the most significant activities. The emergence of LSTM, GRU, and other deep learning technologies models have greatly improved forecasting accuracy. These deep learning techniques, however, are difficult to interpret and analyze which the makes the decision-making process in finance opaque. This research investigates the application of Explanatory AI techniques for improving models interpretability while still maintaining prediction accuracy. The study focuses on attention, saliency maps, Hapley Additive Explanations (SHAP), and Local Interpretable Model Agnostic Explanations (LIME) to determine importance of features for accurate prediction of market trends. It also aims to bridge a gap between explainable deep learning models (LSTM with attention, GRU with attention, and Transformer) and traditional models (LSTM, GRU) by conducting a comparison using financial time series datasets from SP500 and NASDAQ (2010-2024). For this purpose, the study will measure prediction accuracy using MAPE, RMSE, R Squared, as well as training time, all in the context of accuracy and interpretability trade-offs. From the data, we can see that the accuracy of the Transformer models was the highest ,whereas the LSTM + Attention models were more accurate and efficient, and therefore more appropriate for real time use cases. Besides SHAP, the feature importance analysis along with the attention-weighting tools showed market transparency by depicting important market figures. It highlights the primary purpose of XAI concerning compliance regulations, risk management, and AI assisted financial operations. Further studies should be conducted to delve into exploitation of hybrid deep learning models, sentiment oriented ones, and quantum AI in explainable market predictions that facilitate AI integration with industry transparency requirements.

References

G. Box and G. Jenkins, Time Series Analysis: Forecasting and Control, 5th ed. Hoboken, NJ, USA: Wiley, 2019.

T. Bollerslev, "Generalized Autoregressive Conditional Heteroskedasticity," Journal of Econometrics, vol. 31, no. 3, pp. 307-327. DOI: https://doi.org/10.1016/0304-4076(86)90063-1

A. E. Brockwell and R. A. Davis, Introduction to Time Series and Forecasting, 3rd ed. New York, NY, USA: Springer.

S. Hochreiter and J. Schmidhuber, "Long Short-Term Memory," Neural Computation, vol. 9, no. 8, pp. 1735-1780. DOI: https://doi.org/10.1162/neco.1997.9.8.1735

A. Vaswani et al., "Attention Is All You Need," in Advances in Neural Information Processing Systems (NeurIPS), Long Beach, CA, USA, Dec. 2017, pp. 5998-6008.

Z. Zhang et al., "Transformer-Based Financial Forecasting Models: A Comprehensive Study," IEEE Access, vol. 10, pp. 27389-27402, 2022.

S. M. Lundberg and S.-I. Lee, "A Unified Approach to Interpretable Model Predictions," in Advances in Neural Information Processing Systems (NeurIPS), Long Beach, CA, USA, Dec. 2017, pp. 4765-4774.

M. Ribeiro, S. Singh, and C. Guestrin, "Why Should I Trust You? Explaining the Predictions of Any Classifier," in Proc. 22nd ACM SIGKDD Int. Conf. Knowl. Discov. Data Min. (KDD'16), San Francisco, CA, USA, Aug. 2016, pp. 1135-1144. DOI: https://doi.org/10.1145/2939672.2939778

H. Lin, X. Xu, and J. Wang, "Enhancing Market Forecasting with Attention-Based Deep Learning Models," IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 1, pp. 45-58, 2023.

P. R. Gupta and A. Jain, "Sentiment-Driven Attention Mechanisms for Financial Forecasting," Expert Systems with Applications, vol. 213, pp. 118748, 2023.

J. Kim, Y. Kim, and H. Yoon, "Explaining Deep Learning-Based Stock Predictions with SHAP and LIME," Neurocomputing, vol. 512, pp. 163-177, 2023.

B. Zhao, R. Huang, and W. Xu, "A Comparative Study of Explainable AI Methods for Financial Forecasting," IEEE Access, vol. 11, pp. 56789-56802, 2023.

C. Sun et al., "Computational Complexity vs. Model Interpretability in Financial AI Systems," Journal of Computational Intelligence in Finance, vol. 8, no. 2, pp. 142-156, 2023.

N. Patel, A. Sinha, and P. Roy, "Bridging the Gap Between AI Accuracy and Interpretability in Financial Forecasting," Expert Systems with Applications, vol. 215, pp. 119142, 2024.