Towards a New Algorithm for Event Recommendation System

Abstract

We develop a recommendation algorithm for a local entertainment and ticket provider company. The recommender system predicts the score of items, i.e. event, for each user. The special feature of these events, which makes them very different from similar settings, is that they are perishable: each event has a relatively short and specific lifespan. Therefore there is no explicit feedback available for a future event. Moreover, there is a very short description provided for each event and thus the keywords play a more than usual important role in categorizing each event. We provide a hybrid algorithm that utilizes content-based and collaborative filtering recommendations. We also present an axiomatic analysis of our model. These axioms are mostly derived from social choice theory.