In the context of FinTech, we present Financial Graph Attention Networks (FinGAT) to recommend high-profitable stocks in terms of return ratio using time series of stock prices and sector info. Our FinGAT can learn the long- and short-term price tendency, and model the latent interactions or influence between stocks and sectors without any hand-crafted effort. Experiments conducted on Taiwan Stock, S&P 500, and NASDAQ stock markets exhibit remarkable accuracy of FinGAT, comparing to state-of-the-arts (by 11％, 14％, and 12％ performance improvement). The tech is published in IEEE TKDE 2021.

Our tech FinGAT is a graph neural network-based high-profitable financial recommender system. FinGAT can learn long- and short-term tendency of stock prices, and simultaneous capture how stocks and their corresponding sectors are interacted and correlated with each other without external knowledge on the relationships between listed companies. Experiments conducted on TWSE, S&P500, and NASDAQ show that FinGAT outperforms state-of-the-art methods by at least 10％. When incorporating human financial knowledge to select candidate stocks, FinGAT can further lead to 0.97 ranking accuracy.

FinGAT has been extended to several Taiwan’s financial institutions, including Bank SinoPac and E.SUN Bank, to precisely recommend financial items for customers. The main target of FinGAT is financial institutions in Taiwan and in the world. We expect to bring three-fold economic impacts: (1) increasing the market value of listed companies, (2) boosting the transaction volume and rate of the banking industry, (3) increasing customer profitability and raise consumer willingness. FinGAT is open-sourced, and can be used for e-commerce recommender systems, ad placement, and precision marketing.

線上展網址：
https://tievirtual.twtm.com.tw/iframe/166d1efc-3367-4808-8ddb-87441e653a33?group=23bfb1fa-dd5b-4836-81a1-4a1809b1bae5&lang=en