Research And Application Of Bisecting K-means Algorithm Analysis Based On Financial Customer Signature

With the rapid development of the Internet in recent years, a large number of enterprises began to layout the field of e-commerce, with electronic business platform to carry out the marketing and promotion of products. Information technology develops rapidly with the Internet. Fintech has gradually emerged. The arrival of the era of big data for financial institutions is both a challenge, but also an opportunity.Fintech is not a simple literal to provide financial services through the Internet, this is only on the surface of formality behind the needs of the accumulation of large amounts of data and powerful data processing ability, which also are two key factors of Fintech:big data and cloud calculation. Fintech, which relying on big data and cloud computing, provides customers with some of the Internet financial services. And in this thesis, based on fintech attributes of the brokerage business integrated platform is combined with product sales, consulting services, investment contracts, securities transactions as well as relying on big data and cloud computing. For now, there is not an accurate classification financial platform for brokerage business customers, customer signature is only used for simple description of user information. This thesis will build model based on the user's basic information, asset information, transaction records, activities of the platform tack behavior data, cluster analysis and establish data classification model for customers on the basis of the customer signature, get customer classification, and then for each level of the customer to develop personalized marketing programs, so as to more targeted product marketing and promotion.Clustering algorithm is usually used to get customer layer, and K-means algorithm is the most commonly used data mining algorithms, through in-depth analysis of the K-means algorithm, the author finded that choosing appropriate initial centroid is the key to the implementation of K-means algorithm. In general, the random selection of the center of mass is used to solve the factors of human intervention, but this will lead to different operation of the SSE, and ultimately affect the accuracy and stability of the results.In order to overcome the defects of randomly selected centroid, the American scholar Pang-Ning Tan proposed bisecting K-means algorithm, the basic idea of this algorithm is to split the set of all samples into two clusters, one set is selected from the two according to certain conditions, continuing to split, until the K result set is obtained. According to the experimental results, it is concluded that the bisecting K-means algorithm is less affected by the centroid, and the efficiency and accuracy is much higher than that of the K-means algorithm. This thesis is mainly based on the analysis and application of the bisecting K-means algorithm, customer classification by algorithm, different levels of customers match different risk levels of products, so as to achieve the purpose of distinguishing between customer precision marketing strategies.The main work done in this thesis includes:(1) Establish a unified data center, unified customer data by extraction, classification and through a series of methods to filter the data, so that customer data can meet the requirements of the experiment.(2) Establish a customer signature system, establish a unified customer image index system, through a series of indicators to select customers as the basis for customer clustering analysis.(3) By optimizing the methods of cluster analysis to classify customer data, get customer classification, develop personalized marketing programs, and increase customer conversion rates.Based on awareness for important customer research and the current study on Internet financial e-commerce platform for customers, this thesis on the basis of systematic review of classical literature research, will build data modelling through the cloud computing platform with customer's big data. Based on customer signature classify customer by classification algorithm, pinpointing customer and through personalized marketing and promotion to validate and modify data model, improving brokerages customer conversion rates and achieve the expected results.