Research Of Newstock Knowledge Service Based On Web Heterogeneous Information Mining

With the rapid progress of Internet technology, it has affected every aspect of people’s life. The topic of "Internet +" has also promoted the development of a series of industry. In terms of financial knowledge service, Internet platform has become the primary information carrier because of its timeliness of data, convenience to access and friendly visualization in display. With the restart of IPO at 2014, the market of new stock turn to be hot again. And it shows new characteristics under the rules of new market. This paper analysis new stocks. It uses the reliable information for Internet. And it analyzes the multivariate information, researches the problems which users are most concerned of, builds a knowledge service system based on the web heterogeneous information, and shows prediction result and the similarity of stocks for users. The purpose of the research is to provide convenient and comprehensive reference for investors.The research mainly includes the following aspects:Information collection and pretreatment. Information is the foundation of knowledge service system. Information used in our research includes structured market data, web form and unstructured official announcements file. Different method is used to deal with different data. As for the data extracted from the announcements file, we also calculate the extraction accuracy.Prediction about Yiziban. This paper gives the definition of the phenomenon of Yiziban. Research the lasting days of Yiziban by linear model and ensemble learning methods. The precision can reach to 2.7 days and the similarity of stock sequence is 91%. The data set is established by sliding time window. Two kinds of evaluation methods were used to evaluate the results of the prediction.Prediction about the price trend. The trend of new stock is summarized and the prediction problem is defined as a binary classification problems. Use a variety of classification methods to predict the trend about five trading days after Yiziban finished. Make a summary of different algorisms, and the accuracy of classification can reached to 81%.System construction. After information collection and pretreatment, prediction about Yiziban and the price trend, we get the key information and build the financial knowledge service system. We define the similarity of stocks by its industry and the relationship between stocks is displayed by star graph for the first time. All knowledge is showed to inventors by visualize and clear method.