Reinforcement Learning Driven Trading Algorithm with Optimized Stock Portfolio Management Scheme to Control Financial Risk

In recent years, the application of deep learning techniques in financial markets has achieved significant attention for developing effective investment strategies. Traditional schemes focus on optimizing models to maximize returns but often fail to adapt to the dynamic and unpredictable nature of market conditions. Therefore, managing financial risks while increasing returns remains a critical aspect of asset management. This study introduces a unique portfolio management strategy that targets both long-term investments and short-term returns by combining unsupervised and supervised learning techniques. Historical data from Indian stocks listed on national stock exchanges are utilized, with K-means clustering employed to select appropriate stocks. A joint approach of deep learning and reinforcement learning is then used to assign weights to each selected stock, effectively allocating funds for profitable trading. Additionally, an intelligent trading algorithm is devised to operate on a daily/weekly basis, optimizing returns over short trading periods. The proposed model’s reliability and effectiveness are validated through extensive experimental analysis and comparative assessment against secured funds like fixed deposits and standard portfolios such as Nifty-50. The outcome demonstrated the advantages of the proposed portfolio in achieving superior investment outcomes with 35% higher returns.