Bond Knowledge Service System On Heterogeneous Information

With the increasingly combination of the Internet and finance, network knowledgeservice platform of financial products has been recognized by many investors. Theexistence of text information of a large number of financial products in networkprovides a certain source of data for building a knowledge service platform. Therefore,this paper aims to get accurate documents–the listing announcements from the Internet,and extract basic information of financial bond based on the ontology-based rule setfrom them, and then applied the information to the bond knowledge service platform.This paper is mainly described from the following several aspects:Heterogeneous information acquisition and pre-processing: including bond’sstructured data and unstructured web data’s acquisition and pre-processing; Bond datasources include fixed financial websites and search engines, in the search engine sectionwe put forward the optimization model RDMDRR to get more accurate bonddocuments.Bond heterogeneous Information’s process: use artificial annotated method to buildontology rule base of bond feature by WHISK algorithm, use the adaptive ontology-ruleextraction method to prune the ontology rule set. Then adopt the regularization methodto process the preliminary results, it can further improve the extracted entityinformation.Bond Information’s classification and fusion: According to the different categoriesof bond, this paper adopts rule-based method and machine learning method respectively.According to the characteristic of unbalanced distribution of bond categories, this paperpresents an improved feature weighting method IC_TFIDF based on the traditionalfeature weighting method TFIDF, and applies it to the unbalanced bond classification. Itimproves the recognition rate of minority class information and achieves betterclassification results. Then fuse the results with other bond feature together to form amore complete bond knowledge set.After these three steps of information’s process and fusion, the results will beapplied to bond knowledge service system. The new bond knowledge service platformchanges the old knowledge construction mode. The experiment results show that tosome extent it achieves the purpose of automated knowledge service and it is acceptedextensively by the investors.