Hot Spot Detection And Tracking Based On Internet Financial Heterogeneous Information Mining

In recent years, the speculations of concepts and hot spots in the Shanghai and Shenzhen A-share markets are very frequent, such as the Shanghai-Hong Kong Stock Connect, state-owned enterprise reform, Internet finance, One Belt and One Road which are speculated frequently concepts during the first half of 2015. When a concept or hot spot is popular, the related stocks usually tend to have better market performance in a short to medium term period. How to seize the investment opportunities brought by this kind of concepts or hot spots is a difficult work in current quantitative research. Based on the discovery of the hot spots of the A-share market, the research excavates the stocks and sectors related to the hot spots being speculated in the current markets, tracks the development of the hot spots in the market, and provide valuable market information to investors.The main contents of the research are as follows:The acquisition and pretreatment of the internet financial heterogeneous information are mainly divided into three categories: the first category is real-time market data of the A-shares, consists of the daily prices’ fluctuation and turnover of the A-shares, the second is financial information and sector classification of the A shares, and the third is public opinion data of the stocks, including news data and shares it data. After obtaining these data, the raw unstructured financial data are processed into structured data by further de-noising, optimizing, refining and other treatment, to provide accurate data security for the follow-up construction of the system.The hot spots discovery based on the ranking aggregation algorithms: the hot spots discovery of the A-shares’ stock markets can be transformed into a rank aggregation problem by obtaining data associated with the popularity of the hot spots and calculates the corresponding popularity degree, of which each can generate a popularity degree rank. Aggregating all the ranks, a comprehensive hot spot popularity rank can be generated. The research is mainly focused on an unsupervised rank aggregation algorithm, calculating the hot spots’ popularity from three aspects, hot spots in markets, hot spots in medium and hot spots in news, and then the stocks’ popularity and sectors’ popularity, building up an effective module for hot spot discovery.The hot spots track based on the time series prediction methods: After the implementation of the hot spots discovery module, a time-sequential popularity rank sequence for every stock and every sector. The study task of the research is to predict the future time-sequential popularity rank sequence according to the historical time-sequential popularity rank sequence. The research mainly studied the application of the time series prediction methods on the track of hot spots.In summary, using data mining technology, the research excavates the stocks and sectors related to the hot spots being speculated in the current markets, tracks the development of the hot spots in the market, and provide valuable market information to investors.