Predictive Stock Selection: A Hybrid RF-CNN XGBoost Model Integrated with Dynamic Adaptive Index and Stepwise Elimination Techniques

This paper presents a predictive model for stock selection using a hybrid approach. The proposed model, which combines advanced machine learning algorithms with a dynamic adaptive index (DAI), aims to enhance portfolio performance and risk management. In this context, Random Forest (RF) and XGBoost algorithms are utilized to select optimal features and forecast stock returns. In the subsequent step, the DAI method assesses and selects stocks. By integrating various metrics such as correlation, volatility, relative strength index (RSI) and cumulative returns, the DAI method enables a comprehensive assessment of available stocks that dynamically adapts to changing market conditions. In addition, the stepwise elimination technique optimizes the portfolio by eliminating stocks with high volatility and low returns, ensuring that only stocks with strong performance and acceptable risk are retained. The two-stage structure of the proposed model, encompassing stock return prediction and stock selection, enhances the model's overall efficiency. The presented hybrid model excels in stock selection and risk management, helping investors achieve higher returns while reducing portfolio risk.