Nugget Discovery with a Multi-objective Cultural Algorithm
This work addresses the need for interpretable rule mining in data mining for decision-makers, but it is incremental as it applies an existing algorithm type to a specific gap.
The authors tackled the problem of partial classification (nugget discovery) by proposing a multi-objective cultural algorithm to mine comprehensible 'If-Then' rules with user-specified properties, and experimental results on benchmark datasets showed good performance.
Partial classification popularly known as nugget discovery comes under descriptive knowledge discovery. It involves mining rules for a target class of interest. Classification "If-Then" rules are the most sought out by decision makers since they are the most comprehensible form of knowledge mined by data mining techniques. The rules have certain properties namely the rule metrics which are used to evaluate them. Mining rules with user specified properties can be considered as a multi-objective optimization problem since the rules have to satisfy more than one property to be used by the user. Cultural algorithm (CA) with its knowledge sources have been used in solving many optimization problems. However research gap exists in using cultural algorithm for multi-objective optimization of rules. In the current study a multi-objective cultural algorithm is proposed for partial classification. Results of experiments on benchmark data sets reveal good performance.