Web User Profile Inference for User Group Interest Prediction on Social Networks using Domain Ontology

Web inference techniques have become more sophisticated and which can be used for many real world applications like market strategies, business intelligence and etc... In social networks the interest of a single user represents the interest of the whole group to which he belongs to. There exist various approaches to find out the interest of a group, but suffers with the accuracy of clustering users into groups and predicting their interest. We propose a new methodology to predict the interest groups on social networks using domain ontology and their profile details. Each user would be communicating with others in the group or network about some topic and we consider that a single user has N number of interest or topic of conversation. Our ultimate aim is to group similar interested user’s and their interest to infer some valuable knowledge from clustered results. Each user conversation and logs are retrieved to get set of conversation they have with others. From each log the topic of conversation and interest is identified using domain ontology. According to the classes available in the domain ontology the user’s are grouped to form a cluster. The generated clusters are validated by computing overlap measure and the process is iterated till the overlap measure becomes low. The domain ontology has number of classes and each class has different labels which represent the properties and values of domain attributes. The proposed method has produced efficient clusters and the prediction accuracy is also higher.