Unsupervised Bot Detection Using Graph Clustering Technique with Source Code and Docuemnt

Project Discription

Social network is place to connect and share thoughts with each other. But most of people get attract from the social audience gathering for there personal or professional advantages. Social bots are social media accounts controlled completely or in part by computer algorithms. They can generate content automatically and interact with human users, often posing as, or imitating, humans. These are mostly used for advertising, campaigning purposes and to steal users personal data on a large scale. So contributing the data mining research was done in this work by the proposed strategy for characterizing the social user behavior and sorts the bots in the cluster without having any earlier information of the individual tweet / comment. In the propose work no need of any configuration for the information, for example, speakers recognizable proof image or exceptional character, here all procedure is finished by using the diverse mix of content mining field. This work presents a study of methods for detection of user profiles as real or social bot. Here a technique was proposed for classifying the social nodes into two cluster, where digital social network graph was reduce into spanning tree. Here weight of the graph was social sequential action transitional probability. So each node has its own set of transitional probability and distance between nodes of transitional probability act as weight of graph. So as per graph clustering technique spanning tree was developed and highly weighed nodes of this tree act as social bot cluster while other set of nodes are real. Results shows that proposed graph based clustering technique has increase the precision value by 19.59% as compared to previous approach used in [29]. While recall value was also increase by 26.88%, at the same time accuracy of the social bot identification was also increase by 46.53%.

Project Sample Image