GENERALIZED METRIC FOR COMPARING RECOMMENDATION ALGORITHMS

The paper discusses the problem of creating a composite indicator for assessing the effectiveness of recommendation system algorithms. This indicator was developed by combining several metrics using the entropy method. The study is based on testing a set of 12 algorithms and 3 datasets. Foreach combination, several criteria used in practice for evaluating recommendation systems were applied. The results indicate that the composite indicator is a useful tool for assessing algorithm performance. It was found that algorithm quality varies depending on dataset size and other characteristics. The generalized measure can be used to improve algorithm efficiency, develop ensembles, and optimize hyperparameters to enhance recommendation quality.