In recent years, the development of financial markets and the increasing complexity of investor behavior have intensified the need for advanced analytical methods to evaluate the performance of financial instruments. Among these instruments, investment funds play a crucial role in the efficient allocation and mobilization of financial resources within capital markets. Given the high volatility of capital markets and the limitations of traditional methods in capturing nonlinear and dynamic relationships, machine learning approaches—particularly artificial neural networks have gained significant attention for forecasting and performance evaluation purposes.
The objective of this study is to develop a hybrid model based on deep neural networks (DNNs) and capital market time-series data to evaluate the performance of investment funds and predict their returns. In this regard, fund unit returns, the overall stock market index, exchange rates, risk-free interest rates, and macroeconomic variables are used as input features. The research adopts a descriptive–applied methodology, and the proposed model is trained and tested using real data from the Iranian capital market over a multi-year period.
The empirical results indicate that neural network models outperform traditional approaches such as linear regression and classical performance evaluation measures (e.g., Sharpe and Treynor ratios) in identifying nonlinear patterns and predicting fund behavior. Furthermore, the findings suggest that incorporating macroeconomic variables alongside market data significantly improves the predictive accuracy of the model.Overall, the results demonstrate that deep learning-based models can substantially assist investors and fund managers in making more informed decisions and improving risk management efficiency.