Research Of Clustering Algorithm And Application On Automobile Insurance Industry

Data mining is the procedure of extracting implicit, meaningfulinformation and knowledge from data set. Clustering is one of the mostimportant technologies of data mining, and is used to discover unknownclassification in data set. It can also combine with other data mining methodsfor further research to find more knowledge. Clustering has been widely used invariety of field such as commerce, biology, geography, network application andso on. So investigation of clustering algorithm has important meaning in theoryand practice.With the development of information system in insurance industry, thereare much data about automobile insurance policy for business enterprise. Howto apply the data mining technology into the database of automobile insurance,and acquire knowledge from the data resource, becomes a hot topic in theinformatization of insurance industry.Firstly, this paper carries a deep research of K-means which is one ofclustering analysis algorithms, and proposes an efficient clustering algorithm(ECA) based upon vector space initial clustering and nearby cluster search,which is composed of the two sub algorithms SPIC and CANC. On the basis ofspace distribution of the data sample, SPIC algorithm carries out initial clusters,which are then optimized by the CANC algorithm. Stable and efficientclustering is obtained through rational space subdivision by SPIC, avoidobtaining unstable clusters and converging to local optimization. The CANCalgorithm, which is based upon efficiently nearby clusters comparison toremove data point, make full use of achievement of initial clustering, as well asthe stabilization of groups, to reduce the amount of calculation and speed upconstringency. Performance comparison has been carried out between ECA andK-means algorithm through a set of experiments, and the improved algorithm isproved to be feasible and effective.After making an intensive study of the data about automobile insurance,the improved algorithm is then applied to the industry of automobile insuranceto carry out evidence-based analysis. There are two subjects about clusteringanalysis in this paper: customer segmentation and character partition of insurance product purchase. The achievement should be the tacticsimplementation guidance of di fferentiated services, and offer technicalsupport for customer analysis and marketing st rat egies.