International Journal of Information Technology
Publication Type: Journal Article
Author: Mansi Goel; Adwitiya Sinha
Abstract: Human sentiment prevalent in online dialogues largely governs the formation of communities in online social networks. These recommendation methods helps in filtering information to assist in predicting the user activity and participation. However, the processing of such colossal amount of online interactions becomes a herculean task. To address this, we have proposed a nature-inspired recommendation algorithm based on Ant Colony optimization that can efficiently execute over large-scale social interactions. For our experimentation, we have considered a social network of users who actively participate in rating movies. The edge-weight amongst a pair of users denote the measure of similarity of rating behaviour. Our focus is to design a recommender system that could detect the optimal path of propagation using minimally connected dissimilar pairs of users. Our proposed ACO-based Recommendation Method (ACO-RM) is applied to each community of users for finding the shortest path amongst users having least similarities. A shortest path will allow the viewers to receive recommendations in a short period of time. The results obtained from the comparison of proposed ACO-RM with genetic and backtracking algorithm endorsed better performance of our approach in terms of execution time and path cost.
