The complex nature of stock markets, influenced by numerous factors, makes them a fascinating subject for econometric analysis and investors, especially in predicting stock prices which is akin to projecting a company's growth and profitability. With the integration of AI into finance, the use of dynamic Machine Learning (ML) methods for forecasting stock prices has gained renewed interest. This paper introduces a unique hybrid Deep Learning model, CNN-BiLSTM-GRU, enhanced with Phase Space Reconstruction (PSR) technique, a fundamental of the Chaos Theory, aimed at improving the Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) performance metrics in comparison to single models, specifically applied to retail stocks like TGT from Target Inc., AMZN from Amazon Inc., and WMT from Walmart Inc. The findings demonstrate that CNN-BiLSTM-GRU surpasses other models, with PSR contributing to an average enhancement of 37.2% in MAE and 35.1% in RMSE across all datasets. This indicates the significance of applying Chaos Theory to decode the complex patterns in US Retail Stocks. The projects' primary contribution lies in advancing the development of future algorithmic trading strategies through superior sequential time series data analysis.

Presented by:

• Nguyen Quoc Anh – s3926339