Study Of Evaluation Of Effectiveness Evaluation On Validity Clustering

With the advances in technology, the popularity of the computer, the rapiddevelopment of information technology, the wide application of the Internet andthe explosive growth the amount of data, the importance of information isincreasing. People want to obtain valuable information from data, and discoverthe rules and trends of the data. It is impossible to achieve the desired results byobservations and sample calculations, so more effective methods of data analysisare urgent. Data mining is the process of discovering patterns in large data sets,clustering analysis is an important research of data mining direction and mainlyused on pattern recognition, data analysis, image processing, and market research.This work firstly discusses cluster analysis and the evaluation function ofeffectiveness of cluster analysis, tells strengths and weaknesses in the applicationprocess, and the basic flow of the possibility of cluster analysis is alsosummarized. Secondly, based on the possibility theory, we improve the clusteranalysis and the effectiveness of evaluation. The improved cluster analysiscombines the possibility measure with the necessity measure of the probabilitytheory. We use the degree of possibility calculated as the indicator of validitycluster analysis, which solves the repeated clustering phenomenon in theclustering process.The improved function of clustering validity evaluation is mainly composedwith indicators of compactness measure and separability measure. We strengthenthe use of the distance function in compactness measure, and the distancethreshold L is set to mutual restraint with the original some membershipthreshold T and to reduce the amount of noise deleted so that we can avoidinaccuracies due to loss of data caused by the evaluation results. The last, weverify the validity and applicability of the method through the China Mobile’schurn empirical analysis.