Application Of Data Mining Technology In The Customer Preference Analysing Of Securities

The thesis describes how to provide good customer service, devote to the innovation of customer classification in eligibility management, establish a scientific and effective classification methodology, and promote it for securities companies in the increasingly competitive securities industry. Index customers' transactions and natural attributes, and choose approprivate platforms and algorithms of data mining to analyze customers' behaviours and classify customers according to the analysis. It helps us to know of customers and setup scientific investment philosophy. Security sales department is able to lead customers' investments according to customer classification, lay a solid foundation for having an in-depth of adequacy service,improve service competitiveness, and strengthen cusomters' loyalty.On the basis of building data warehouse based customer transactions storage and index system, use PCA to select features, use K-means to model for customer perferences, cluster securities customers' perferences of products and transaction opportunities, and re-strip clustering results.Customer perference model selects customer information and customer transactions as the basis index system. Build the model with data mining based on data warehouse.In terms of implementation, system selects three months of customer behavior data from history library at the beginning of 2011, including client mandates,customer transactions, customer funds, daily positioning details, daily tradingsettlement funds, securities market information, securities information, customer basic information. Analyze customer preferences via system cluster method(hierarchical clustering), non-hierarchical cluster method. Comparing these two kinds of algorithms and techniques, choose K-means as the final model and prove that the algorithm is effective.The model was proven to be effective by the validation of the customers' behavior data. The product data of one securities company shows that the model has practical significance on improving the accuracy and effectiveness of clustering customers' perferences.