Predicting Behavioral Analysis of Social Network Users Using Constrained Clusters Based Recommendations

Abstract

Online Social Networks (OSN) has come across an exponential increase in sharing newlinedata by considering both user frequency and information frequency in the network. newlineIn OSN, information is shared between users and the overall network, to achieve newlineinformation or services for specific objects or items in a specific domain. Social users newlinemake a search on the search engines according to their queries. While analyzing newlineand grouping users, there commendation is playing a major role in OSN. Recently, newlineseveral machine learning and recommendation mechanisms have been designed for newlineOSNs, with higher recommendation accuracy. While improving the accuracy of the newlinerecommendations, the running time of user grouping is high and failed to consider the newlinefalse positive rate. newlineTo overcome the above limitations in this work, three different techniques are newlineproposed to improve recommendations accuracy and communication overhead with newlineminimum error rate. This work introduces three different methods namely Map Reducebased newlineWeighted Multi-constraint Correlated Clustering (MR-WMCC), Optimized newlineExtreme Machine Learning and Orthogonally Projected (OEML-OP) and Soft Cosine newlineGradient and Gaussian Joint Probability (SCG-GJP). These techniques are used for newlinegrouping similar users on the online social networks. newlineMap Reduce-basedWeighted Multi-constraint Correlated Clustering (MR-WMCC) newlinemethod is proposed initially for identifying the group of users in OSN. The MR-WMCC newlinemethod is designed with the process of Map Reduce-based Tag Identification (MRTI) newlinemodel and Weighted Multi-constraint Correlated Clustering (WMCC) model. At first, newlinethe MRTI model is carried to perform behavior patterns of users according to the actual newlinetweet list. Here, user communication between the OSN users is said to obtain based newlineon the hashtag. Also, parallelism factor is applied for attaining reduced dimensionality. newlineWith the hashtag based on Map Reduce framework using Hadoop, the unwanted string newlineis eliminated and therefore reducing the input tweet dimensionality. Thus, the run newlinetime of user group