A meta-learning approach for recommending a subset of white-box classification algorithms for Moodle datasets.
Cristóbal Romero, Juan Luis Olmo, Sebastián Ventura · Educational Data Mining · 2013
This paper applies meta-learning to recommend the best subset of white-box classification algorithms when using educational datasets. A case study with 32 Moodle datasets was employed that considered not only traditional statistical features, but also complexity and domain specific features. Different classification performance measures and statistics tests were used to rank algorithms. Furthermore, a nearest neighbor approach was used to recommend the subset of algorithms for a new dataset. Our experiments show that the best recommendation results are obtained when all three types of dataset features are used.