Data Mining And Program Framework Design Of Society Label Facing SNS

More and more people participate in the edition of internet content, and folksonomy gradually becomes one significant way of classifying the internet resource. This paper mainly studies the advantages and disadvantages of folksonomy and particularly studies one of the disadvantages-ambiguity. The paper proposes the improved clustering scheme, which is based on analyzing the cause of ambiguity and studying others'academic results. What's more, the paper designs program of crawl and clustering. During the process of designing the framework, the paper divides the whole program into four segments:data crawl, cleaning data, clustering and image demonstration. Specifically speaking, the paper achieves the whole framework by mixed programming with VC++6.0 and SQL2005, which combines the four parts comprehensively by the designing mode, and the framework is low coupling high cohesion, easy to broaden and to maintain. Lastly, the paper makes a black box test whether the framework is correct, the database is visited correctly and the clustering results are reasonable. In addition, the paper analyzes the aggregation which is based on the semantic clustering scheme and judges whether the anticipated purpose is meet. Furthermore, the paper analyzes the cause of forming different aggregation when the clustering conditions are changed many times. The paper studies the label of folksonomy and generates new aggregation by combing the label again, which means classifying professionally the folksonomy in a second time to form new label group. This way reduces the ambiguity of label, meanwhile it improves the check all rates, hit rate and order precision when the internet users inquire.