Developing optimization models for dyadic prediction in the context of sparse data

Abstract

There is renewed interest in recommender systems (RS) research after the Netflix prize competition. in parallel, there is also a growth of social medial which has driven research in link prediction, with the aim of predicting lables about friends, groups, etc. Both these problems involve the prediction of labels (star ratings or friendship) between a pair of entities (user-movie or user-user). This type of problem is known as a dyadic prediction. Dyadic prediction in context of 'personalization'on e-service platforms is a broad research area of this dissertation. In particular, this dissertation seeks to contribute three folds on the implementation of personalization tools on e-commerce.