معین محمدزاده فردفینی

Stock price forecasting is one of the key challenges in the financial and investment field, which is very important due to the unpredictable fluctuations of the market and the influence of various economic, political and social factors. With the increase in the amount of available data and the variety of information, the use of advanced models to analyze these data and identify patterns affecting prices can help improve investors' decisions. The purpose of this research is to investigate the effect of multi-source data and deep learning models on stock price prediction and optimal investment portfolio design. Data were extracted from various sources including financial databases, online financial news and technical analysis. This research was carried out in three different scenarios: the first scenario, stock price prediction was done using raw price data only; The second scenario, forecasting was done by combining price data and technical indicators; And the third scenario, the prediction was made using raw price data, technical indicators and polarity analysis of financial news. The results of this research showed that the combination of multi-source data significantly increased the accuracy of stock price prediction, and deep learning models, especially BiRNN and BiLSTM, provided the best performance in this field. Also, interpretable artificial intelligence tools such as SHAP and LIME helped to identify the factors that have the greatest impact on price changes and made it possible to understand the complex relationships between input characteristics and forecasting results. Finally, using the Markowitz theory and Sharpe ratio, the optimal investment portfolio was designed, which obtained the expected annual return of ۶۰.۳۹% and the risk of ۱۵.۱۳%. This research is known as a key step in improving the processes of stock price forecasting and investment portfolio optimization, and it can be a good basis for future research in this field. According to the obtained results, the importance of using new approaches based on artificial intelligence and multi-source data in investment decisions and risk management is emphasized.