Combining Aspects of Genetic Algorithms with Weighted Recommender Hybridization
This work addresses the need for improved recommendation quality in systems like movie or name recommenders, though it is incremental as it builds on existing hybridization methods.
The paper tackles the problem of combining unscored recommendations from multiple algorithms into a single ensemble, proposing a method inspired by genetic algorithms. It outperforms a weighted voting method by 20.3% and 31.1% on movie- and name-recommendation datasets and reduces execution time by up to 19.9%.
Recommender systems are established means to inspire users to watch interesting movies, discover baby names, or read books. The recommendation quality further improves by combining the results of multiple recommendation algorithms using hybridization methods. In this paper, we focus on the task of combining unscored recommendations into a single ensemble. Our proposed method is inspired by genetic algorithms. It repeatedly selects items from the recommendations to create a population of items that will be used for the final ensemble. We compare our method with a weighted voting method and test the performance of both in a movie- and name-recommendation scenario. We were able to outperform the weighted method on both datasets by 20.3 % and 31.1 % and decreased the overall execution time by up to 19.9 %. Our results do not only propose a new kind of hybridization method, but introduce the field of recommender hybridization to further work with genetic algorithms.