| With the development of information technology, the data resources of enterprise isgrowing explosively. The traditional competitive intelligence system exposes itsdeficiencies of the data collection and processing. However, the rise of data miningprovides a new impetus for Enterprise Competitive Intelligence System. Fuzzy clusteranalysis is an important research topic in data mining and widely applied in variousfields. Therefore, the research on fuzzy cluster analysis in the Enterprise CompetitiveIntelligence System is of great significance.With the application in Enterprise Competitive Intelligence System, this papermade the following improvements for the problems of FCM (fuzzy C means algorithm):(1) For the problem that FCM algorithm can’t get the best number of clusters and isso sensitive to its initialization that could be easy to fall into local optimization, thispaper proposed an improved algorithm which is based on K means and principle ofgranularity. At the same time, when use effectiveness evaluation function to determinethe best number of clusters, different number of clusters would select initial clusteringcenter for FCM clustering again, which would make the best number of clustersunstable. As a result, we proposed a method of merging cluster centers.(2) For the problem that characteristics of each dimension has differentcontributions in course of clustering, this paper proposed an improved FCM algorithmwhich is based on features weighted. The algorithm extracted main feature vectors usingprincipal component analysis, and gave corresponding objective weight. Besides, itfeature weighted sample date according to the given subjective weight, and clustered theweighted data using the improved FCM algorithm which is based on K means and theprinciple of granularity.Finally, based on the previous research and according to projects before, weintroduced an improved FCM algorithm to the Enterprise Competitive IntelligenceSystem, which achieved better results. |
PDF Full Text Request