The Research Of Character Relationship Analysis And Visualization

In the Internet era,the network activities that people participate in are more frequent and widespread.we can process and analyze a large number of user attribute information and behavior data by using classification analysis,clustering analysis and association analysis technology in data mining,and find out some potential rules and relevance between the users,then mine relationships through data and represent the results with visualization methods,it is a hot topic in many fields.The paper's work was based on clustering analysis,six degree segmentation theory,centrality algorithm and association analysis technology to excavate and analyze the attribute information of characters,so as to discover the implicit relationship among the characters,and then displayed and analyzed the visualization to the relational network..The main research contents and innovation points are as follows:Firstly,the traditional K-modes clustering algorithm did not determine the influence of redundant data on clustering results,nor could it determine the weight values of attributes.Aiming at these problems,the K-modes clustering algorithm based on rough set and information entropy was proposed.It combined the rough set theory and information entropy theory to the traditional K-modes clustering algorithm,removed redundant attributes by using rough set attribute reduction algorithm,extracted the necessary attributes,and combined with the concept of information gain,to further determine the attribute weights,which improved the algorithm efficiency and accuracy.Finally,the performance of the pre-improved K-modes and after-improved K-modes algorithm were tested in five data sets of the UCI machine learning library,such as soybean-small,zoo and so on,and the class accuracy and classification accuracy were analyzed as well.The experimental comparison showed that the performance of the K-modes clustering algorithm based on rough sets and information entropy was better than the traditional K-modes algorithmThen,according to the character attribute data in this paper,it combined with the six degree segmentation theory and centrality algorithm,gave the definition of key person and key index,and put forward to the method and algorithm which could be applied to mine the key nodes of character relationship.In addition,aiming at the problem of the time complexity of the traditional six degree segmentation algorithm,made the single search in the traditional search algorithm become two-way search,so as to become the time complexity reduced to half of the original one;finally,according to the algorithm and the proposed method,the experiments were carried out on the karate club network,the dolphin network,the American Football League Network and the randomly selected character attribute data sets,and the results were analyzed,and ultimately,improved that,the key personage mining algorithm proposed in this paper was feasible and effective.Lastly,in view of the user's acceptance of information and the observability of information,The visual display and analysis of the relational network were carried out on the actual character data set through the visualization tool--Gephi,it included visualization of key figures excavated and visualization of key person relationship and network,etc.,which made it easy for users to understand objectively whether the relationship between objectives is intimate or not.