Abstrakt: |
Objective With the increasing volume of information and the complexity of financial markets, investors are increasingly seeking innovative financial tools to make more informed decisions. These tools should help investors choose the right stocks and achieve better returns. In this regard, stock recommendation systems are becoming increasingly important. Stock recommendation systems can assist investors in achieving superior returns by selecting the right stocks. However, traditional stock recommendation systems often lack the necessary accuracy and efficiency. This research aims to develop a novel approach called Stock-based Collaborative Filtering to design a stock recommendation system for the Tehran Stock Exchange. This method is founded on two key assumptions: first, market inefficiency, meaning the stock market does not completely and accurately reflect all available information; and second, the presence of hidden information in stock movements, indicating that these movements contain valuable insights that can influence the prices of other stocks in the market. In this research, assuming the existence of the transmission effect on the Tehran Stock Exchange, we used the collaborative filtering technique, a common algorithm in recommender systems, to design a stock recommender system. The purpose is to help investors select the best-performing stocks to outperform the market. Methods This study uses historical stock price data of 145 firms listed on the Tehran Stock Exchange from 2012 to 2021. The collaborative filtering algorithm was implemented in two stages: training and testing. In the training stage, the algorithm was trained using data from 2012 to 2016, and in the testing stage, its performance was evaluated on data from 2016 to 2021. Following, buy and sell signals were generated using the stock-based collaborative filtering algorithm during the same period. Finally, the strategy was evaluated. Results The algorithm was tested as an investment strategy in both in-sample and out-of-sample periods. The results obtained from the algorithm for the out-of-sample periods show that this strategy can achieve a 25-fold return. The overall index returned 16 times during this period, indicating the excellent performance of the strategy over time. Additionally, the value at risk (VaR) for the selected method during the study period stood at -12.8%, indicating the lower risk of this method. Conclusion Stock-based collaborative filtering is an active investment strategy. This intelligent algorithm aims to identify undervalued stocks and achieve higher returns than the market. This algorithm can serve as a valuable tool for active investors seeking to identify valuable stocks and achieve higher returns than the market. Therefore, further research is necessary to examine the performance of this algorithm in different markets and diverse economic conditions. Also, it is recommended to implement risk control strategies and optimize the system's efficacy further. [ABSTRACT FROM AUTHOR] |