Online Social Voting Using Collaborative Filtering Based Method
Loading...
Files
Date
2020-09-24T11:24:34Z
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Social voting is becoming the new reason behind social recommendation these days. It helps
in providing accurate recommendations with the help of factors like social trust etc. Here we
propose Matrix factorization (MF) and nearest neighbor-based recommender systems
accommodating the factors of user activities and also compared them with the peer reviewers,
to provide a accurate recommendation. Through experiments we realized that the affiliation
factors are very much needed for improving the accuracy of the recommender systems. This
information helps us to overcome the cold start problem of the recommendation system and
also y the analysis this information was much useful to cold users than to heavy users. In our
experiments simple neighborhood model outperform the computerized matrix factorization
models in the hot voting and non hot voting recommendation. We also proposed a hybrid
recommender system producing a top-k recommendation inculcating different single
approaches.These ratings are given based on a individual’s voting or opinion. But if we
consider a case as an example where only a single user or a less number of users have given
the review as good or above average, the overall review delivered would be a good or above
average because only some users have rated it. In this way an accurate prediction is not
delivered to the end user. So, this paper proposes a hybrid collaborative system, which
calculates the movie overall review by comparing the individual’s review based on previous
activities, based on the comparison of the users with the others who gave similar ratings and
compares the individual ratings to all other people’s rating for the very same movie and ranks
it based on Top 10, Top 20 and Top 50. The review is based on hot voting and cold voting
where the hot voting is based on the user’s participation in giving the rating and a cold user
who rarely gives the review. We show that simple meta path-based NN models outperform
computation-intensive MF models in hot-voting recommendation, while users’ interests for
nonhot voting can be better mined by MF models. Also, this paper proposes a method to know
whether a user is likely to watch a movie or not based on the k-nearest neighbor algorithm.
Description
Keywords
VANISHA P, 1NH17MCA54