LEONARDO - The computational intelligence (CI) model selection wizard

Abstract

The need for tools to aid the selection of the CI models that lie at the heart of many AI systems has never been greater, due to the mainstreaming of data mining and other AI applications. LEONARDO -our contribution to this process- is a recommender system that selects and ranks applicable CI models for a given problem based on the peculiarities of the domain as determined by the user's preferences and dataset characteristics. Leonardo's recommendations are based on two knowledge bases. One contains the description of 65 CI models and provides the Meta knowledge for pruning the space of all CI models to only those applicable to the current task. The second KB contains the performance results of over 200 datasets on the applicable CI models. LEONARDO's ranking is achieved by using the performance information of the k entries, from this KB, nearest in similarity to the new domain dataset.