Abstract
Stock trend prediction plays a critical role in seeking maximized profit fromstock investment. However, precise trend prediction is very difficult since thehighly volatile and non-stationary nature of stock market. Explodinginformation on Internet together with advancing development of natural languageprocessing and text mining techniques have enable investors to unveil markettrends and volatility from online content. Unfortunately, the quality,trustworthiness and comprehensiveness of online content related to stock marketvaries drastically, and a large portion consists of the low-quality news,comments, or even rumors. To address this challenge, we imitate the learningprocess of human beings facing such chaotic online news, driven by threeprinciples: sequential content dependency, diverse influence, and effective andefficient learning. In this paper, to capture the first two principles, wedesigned a Hybrid Attention Networks to predict the stock trend based on thesequence of recent related news. Moreover, we apply the self-paced learningmechanism to imitate the third principle. Extensive experiments on real-worldstock market data demonstrate the effectiveness of our approach.