A Stock-Selection Model Using Text Mining,Portfolio Modeling And Diversification Research

Text mining is a comprehensive subject, involving a series of Technologies like mathematics, statistics and computer science and so on, which is a powerful weapon to deal with the information explosion. The information on the Internet is beyond count, and surely there is useful knowledge, but more noise. Only with the rapidly developed modern computer, we can use the distributed computing tools such as Hadoop and Map-Reduce to process big data.Bi Bin reviews the origin and development of Asset Portfolio Theory (APT), its profound influence in the capital market, and also the challenges faced by traditional finance theory coming from the anomalies. So some scholars draw lessons from the theories of social science, behavioral science and psychology, etc. They give explanation to the anomalies and develop behavioral portfolio theory.On the basis of previous research, Bi Bin proves the feasibility of the implementation of the text mining system. He makes use of computer technology to analyze financial text sentiment, builds a text mining system, aiming at analyzing the financial news and ’xueqiu.com’ postings. He uses the web crawler to download the webpages, conducts text classification with the distributed computing technology and large-scale support vector machine (LSSVM). He designs two indicators:stock attention degree and quantitative emotion degree, quantitatively analyzing the financial text data. He combines these two indicators into the traditional value stock-selection model to establish the text mining stock-selection model.Bi Bin creatively establishes a Chinese Text Mining System (CTMS), the first investor-social-platform-oriented one. He designs a buy/sell signal with the text mining stock-selection factor, and proves the validity of the signal. To compare the traditional stock-selection model and the text mining stock-selection model, he introduces’Effective Number of Bets (ENB)’to estimate the risk parity level of the two models. Through the comparison he finds that adding text mining factor into the traditional model, one can get a new perspective different from corporate finance and stock volume/price factor. Text mining factor effectively increases the diversification level, so the portfolio is expected to get a low risk and stable return. The effectiveness of text mining model also provides evidence of the existence of the investment behavior bias. Through the identification of the deviation, one can possibly gain excess returns from the text mining stock-selection model.Through this study, Bi Bin deepened the understanding of modern portfolio theory, and he conducts very useful research of the development of behavioral finance and the application of text mining technology.