Mining gold in senior executives´ pockets: An online automatically trading strategy

Online financial news is an important part of financial Big Data. In this paper, we propose a model to promptly recognize valuable news about senior executives´ behavior and an online automatically trading strategy based on the model. Our model consists of three phases. First, word segmentation and keyword extraction are employed to quantify the financial text. For a better efficiency and promptness, manifold learning is utilized to reduce the dimension of keyword vector. Second, the idea of financial event study is utilized to judge whether a specific type of news could produce significantly positive or negative return. Third, support vector machine is employed to recognize the specific financial news and associate the quantified text with the stock return. Experiments show that the recognition work performed excellently and the behavior of increasing shareholdings produces significant positive return. Our online automatically trading strategy based on the model obtained a return of 55.62%, outperforming three main benchmarks in the same period, 4.52%, 12.47% and -6.89% respectively.