Please use this identifier to cite or link to this item:
|Title:||Friend/followee recommendation system for users based on interests : advanced AI algorithms||Authors:||Ailesh Sethu Piramanayagam||Keywords:||DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence||Issue Date:||2012||Abstract:||Interest graph, a mapping of people and their relationships based on their interests, is a popularized concept in the technology industry over the last year or so. It provides quality content for the social network's users and a more effective way for advertisers to target audience groups compared to the prevalent methods in social networks like Facebook, Twitter etc. Collaborative filtering is a technique used to predict items of interest for an active user based on the level of similarity of the user with other users in the data set or items' similarity with one another. Pearson's product-moment coefficient is the most common way of depicting correlation between two random variables. In the case of user based collaborative filtering, the correlation between users is used to predict products the user might like. Taking inspiration from that idea, other users similar to the active user can be suggested as Follow/Friend Recommendations in a social network. Kohonen's Self Organizing Maps helps in clustering similar items in higher dimensional space by reducing them to lower dimensions, in most cases two dimensional space. Among many similarity measure for Self Organizing Maps, the Euclidean Distance between two nodes is one of the most common ways of depicting similarity between two nodes in the given dimension. While current systems proposed for Follow/Friend Recommendations is based on social proximity and collaborative filtering, little work has been done in the field of neural networks being implemented for recommending similar users. Through this project, the author wishes to implement the existing Self Organizing Maps concept to the problem of Follow/Friend Recommendations and generate a system to analyse the performance of the implemented neural network against the traditional Collaborative Filtering model based on Pearson's correlation coefficient.||URI:||http://hdl.handle.net/10356/49142||Rights:||Nanyang Technological University||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||SCSE Student Reports (FYP/IA/PA/PI)|
Page view(s) 10289
checked on Oct 28, 2020
checked on Oct 28, 2020
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.