Portfolio Optimization with Return Prediction: A Deep Learning and Machine Learning Approach for Cryptocurrency Markets

Constructing an optimal portfolio is a key concern for traders and investors. Since the Markowitz mean-variance model, extensive research has refined portfolio optimization. Recently, machine learning and deep learning have shown promising results in return prediction. This study analyzes four years of data for the ۵۰ largest cryptocurrencies. Using Support Vector Regression (SVR) and Convolutional Neural Networks (CNN), it predicts daily returns for the next year. The models are evaluated using Mean Squared Error (MSE) and Mean Absolute Error (MAE). Predicted returns are integrated into the Mean-Variance-Forecasting (MVF) model, optimizing portfolio allocation while considering forecast error. The MVF model’s performance is compared to the traditional mean-variance (MZ) model, which ignores errors. Finally, four models are analyzed, showing that CNN outperforms SVR in return prediction. The results highlight deep learning’s potential for financial forecasting and portfolio optimization.

کلیدواژه ها:

Financial Trading ، Return Prediction ، Portfolio Optimization ، Deep Learning ، Machine Learning