Experimental Evaluation of the e-LICO Meta-Miner

Operator selection is the task of selecting the right operator for building not only valid but also optimal data mining (DM) workflows in order to solve a new learning problem. One of the main achievements of the EU-FP7 e-LICO project has been to develop an Intelligent Data-Mining Assistant (IDA) to assist the DM user in the construction of such DM workflows following a cooperative AI-planning approach [2] coupled with a new meta-learning approach for mining past DM experiments, referred as the e-LICO meta-miner [3]. The idea of meta-mining [1] is to build meta-mined models from the full knowledge discovery process by analysing learning problems and algorithms in terms of their characteristics and core components within a declarative representation of the DM process, the Data Mining OPtimization ontology (DMOP). In this paper, we provide experimental results to validate the eLICO meta-miner’s approach to the operator selection task. We experimented on a collection of real-world datasets with feature selection and classification workflows, comparing our tool with a default strategy based on the popularity of DM workflows. The results show the validity of our approach; in particular, that our selection approach allows to rank appropriately DM workflows with respect to the input learning problem. In the next section, we briefly review the metaminer. In section 3, we present our results. And in section 4, we conclude.