Analysis of Data Models for Recommender Systems using Factorization Machines
Music recommender systems aim to utilize different kinds of data to predict tracks that a user might like. To achieve this goal, factorization models have gained lot of popularity in recent years. This master thesis facilitates Factorization Machines, a machine-learning approach that can incorporate
contextual data into the underlying model while computing predictions. The aim is to perform an in-depth analysis of different data models (potentially including contextual data) regarding their impact and influence on the performance of the recommender systems. Furthermore, we are also interested in tuning the underlying Factorization Machine towards the developed data models. The evaluations conducted are based on a previously gathered, pre-compiled dataset.