comparison of the portfolio optimization methods based on historical return approach and predicted return using arima model

Risk assessment and the selection of an optimal portfolio selection are critical issues in financial research, investment firms, and among investors. traditional optimization models like mean-variance, value-at-risk, conditional value-at-risk, and omega often rely on historical returns, which can be insufficient in optimizing return and reducing risk. recently, researches have used predictive return models in the optimization process. this study uses arima model to predict stock returns alongside a mean-variance optimization model (arima_mv) using iranian exchange market data from 1395 to 1401. first, stock returns have been predicted using arima model and error criteria such as mse, mad, hr, hr+ and hr- are estimated. the results show that despite its simplicity, the arima model has a relatively good performance in predicting the return of stock. then, using the predicted return and the mean variance model, the optimal portfolio has been calculated in the sliding window process. the results show that portfolios calculated by arima_mv model outperform the tehran exchange index (tedpix), risk-adjusted criterion like sharpe and jensen’s alpha ratios and mean_variance model with historical data (hmv). with the model developed in this project (arima_mv), investment companies can offer shareholders higher returns and lower risks, potentially increasing company value in the capital market and boosting shareholder wealth.