Personalized Stock Recommendations Focusing on Temporal Preferences and Diversification Effect

Abstract

Recommender systems are vital in complex financial markets, helping individuals make well-informed decisions. While many studies have focused on predicting stock prices, the reality is that even advanced models fall short of accurately forecasting them. Additionally, previous studies indicate that individual investors often disregard established investment theories, favoring their personal preferences instead. This presents a challenge for stock recommendation systems, which must not only provide strong investment performance but also respect these individual preferences. To create effective stock recommender systems, three critical elements must be incorporated: 1) individual preferences, 2) portfolio diversification, and 3) the temporal dynamics of stock features and investor preferences. In response, we develop our new model, the temporal graph network approach with mean-variance efficient learning, which can handle time-varying collaborative signals and incorporates diversification-enhancing sampling. Running our model on real individual trading data from the Greece stock market, our approach demonstrated superior performance compared to various baselines, including cutting-edge dynamic embedding models and existing stock recommendation models, showing better investment performance while effectively capturing individual preferences. Furthermore, through user case studies conducted post-experiment, we analyzed the sectors and momentum of the purchased stocks and recommended stocks. We compared these analyses across different models to meticulously examine in which aspects our model can provide better stock recommendations.