RECOMMENDER SYSTEMS BASED ON ONLINE SOCIAL NETWORKS -AN IMPLICIT SOCIAL TRUST AND SENTIMENT ANALYSIS APPROACH

Access to files

•   FULL-TEXT.PDF (pdf)

Abstract

Recommender systems (RSs) provide personalised suggestions of information orproducts relevant to user’s needs. RSs are considered as powerful tools that help usersto find interesting items matching their own taste. Although RSs have made substantialprogress in theory and algorithm development and have achieved many commercialsuccesses, how to utilise the widely available information on Online Social Networks(OSNs) has largely been overlooked. Noticing this gap in existing research on RSsand taking into account a user’s selection being greatly influenced by his/her trustedfriends and their opinions, this thesis proposes a novel personalised RecommenderSystem framework, so-called Implicit Social Trust and Sentiment (ISTS) based RSs.The main motivation was to overcome the overlooked use of OSNs in RecommenderSystems and to utilises the widely available information from such networks. Thiswork also designs solutions to a number of challenges inherent to the RSs domain,such as accuracy, cold-start, diversity and coverage.ISTS improves the existing recommendation approaches by exploring a new sourceof data from friends’ short posts in microbloggings. In the case of new users who haveno previous preferences, ISTS maps the suggested recommendations into numericalrating scales by applying the three main components. The first component is measuringthe implicit trust between friends based on their intercommunication activities andbehaviour. Owing to the need to adapt friends’ opinions, the implicit social trust modelis designed to include the trusted friends and give them the highest weight of contributionin recommendation encounter. The second component is inferring the sentimentrating to reflect the knowledge behind friends’ short posts, so-called micro-reviews.The sentiment behind micro-reviews is extracted using Sentiment Analysis (SA) techniques.To achieve the best sentiment representation, our approach considers the specialnatural environment in OSNs brief posts. Two Sentiment Analysis methodologiesare used: a bag of words method and a probabilistic method. The third ISTS componentis identifying the impact degree of friends’ sentiments and their level of trustby using machine learning algorithms. Two types of machine learning algorithms areused: classification models and regressions models. The classification models includeNaive Bayes, Logistic Regression and Decision Trees. Among the three classificationsmodels, Decision Trees show the best Mean absolute error (MAE) at 0.836. SupportVector Regression performed the best among all models at 0.45 of MAE.This thesis also proposes an approach with further improvement over ISTS, namelyHybrid Implicit Social Trust and Sentiment (H-ISTS). The enhanced approach appliesimprovements by optimising trust parameters to identify the impact of the features(re-tweets and followings/followers list) on recommendation results. Unlike the ISTSwhich allocates equal weight to trust features, H-ISTS provides different weights todetermine the different effects of the two trust features. As a result, we found thatH-ISTS improved the MAE to be 0.42 which is based on Support Vector Regression.Further, it increases the number of trust features from two to five features in order toinclude the influence of these features in rating predictions. The integration of the newapproach H-ISTS with a Collaborative Filtering recommender system, in particularmemory-based, is investigated next. Therefore, existing users with a history of ratingscan receive recommendations by fusing their own tastes and their friends’ preferencesusing the two type of memory-based methods: user-based and item-based. H-ISTSitemis the integration of H-ISTS and item-based which provides the lowest error at 0.7091.The experiments show that diversity is better achieved using the H-ISTSuser which isthe integration of H-ISTS and user-based technique.To evaluate the performance of these approaches, two real social datasets are collectedfrom Twitter. To verify the proposed framework, the experiments are conductedand the results are compared against the most relevant baselines which confirmed thatRSs have been successfully improved using OSNs. These enhancements demonstratethe effectiveness and promises of the proposed approach in RSs.

Keyword(s)

Recommender systems, trust, sentiment analysis

Bibliographic metadata

Institutional metadata

Record metadata