An Association Rules-based Six Degree Separation System: Design And Implementation

With the fast development of Internet technology, Web2.0 becomes the hot spot in the world. Blog, Tag, SNS, RSS, Wiki applications are chief representives of Web2.0 in the global scope. In Web2.0 upsurges, SNS (Social Networking Service) website's potential development attracts most attentions. SNS is an online experiment based on the six degrees of separation where individuals can attempt to contact a stranger in another part of the world through their connections which can offer individuals useful sources of people. Six Degrees of Separation System is an application based on SNS website, help users to seek and establish stable six degree relations in the SNS website. And then take commercial use above six relations. At present, foreign and domestic researches on the establishment of six degree relations are almost based on the technology of search engines which match the user by keywords. They can not dig out the intention of users' queries and the socialized demand.In view of practical demand, the paper combines two most important technologies: Association Rules and Classify clustering in data minging domain with the method of weigh into the application of six degree separation. This system works in the client, uses association rules to dig out friend-pattern, combines "the system rating" and "the individual rating" to enhance the accuracy of rules, improves the clustering technology by weighing method. As result, the system provides each user:1. Weighed first-degree relations;2. Classifies the strangers for the user according to the different classified rules, enables the user take commercial activities above the classes, such as sending emails and so on.The paper's primary works are as follow:1. To pre-process the user information, the paper presents several methods based on the six degree relation systems, including text data generalization method , numerical data clustering method and relation based hot deck imputation method;2. Researching on weighed class association rules, the paper presents "the system rating" and "the individual rating" to enhance the accuracy of rules;3. In the implementation of weighed classifying, the paper improved traditional CBA stability by rule-weighing method. Adjusts the intensity of rules using rule weighing means WNew_CBA. The method can improve radically the quality of training set and reach purpose of classification performance; 4. Designs and implements the Association Rules-based six degree separationsystem.Compared with conventional six degree separation functions, Association Rules-based six degree separation system is intelligent and targeted at the user's interest. It not only meets the personalized requirements, but also provides intelligent and personalized searching assistance, and it has a promising application foreground.