Recommender Systems Algorithm Selection Using Machine Learning

This article delivers a methodology for recommender system algorithm selection using a machine learning classifier. Initially, statistical data from real collaborative filtering recommender systems have been collected to form the basis for a synthetic dataset since a real meta dataset doesn't exist. Once the dataset has been developed a classifier can be applied to predict which recommender system among a range of algorithms will predict better for a given dataset. The experimental evaluation shows that tree-based approaches such as Decision Tree and Random Forest work well and provide results with high accuracy and precision. We can conclude that machine learning can be used along with a meta dataset comprised of statistical information in order to predict which recommender system algorithm will provide better recommendations for similar datasets.