Chasing Returns of Open-End Investment Funds Using Recurrent Neural Networks. A Long-Term Study

The primary motivation of this study is to empower individual investors with a data-driven strategy for finding long-term investment returns by leveraging recurrent neural networks (RNNs) to forecast fund performance and construct dynamic portfolios. Specifically, we use RNN to forecast the returns of open-end investment funds and build a portfolio of top-performing funds based on these forecasts. Using a sample of 71 equity, fixed income, hybrid and money market funds in the Polish market from 2005 to 2022, we train the network over five years to generate annual logarithmic return forecasts for each fund. These forecasts underpin a straightforward long-term investment strategy: at the end of each forecasted year, funds with positive returns are added to the portfolio. In subsequent years, the portfolio is adjusted by retaining or adding high-performing funds and removing underperforming ones. Our findings reveal that this strategy delivers higher returns than passive investing or traditional regression-based models, making it a viable long-term option for individual investors aiming to secure their retirement. By showcasing its superiority over conventional methods, the study offers a practical and adaptable solution for achieving financial security in dynamic market environments.