A social intelligent system for multi-objective optimization of classification rules using cultural algorithms

Cultural algorithms (CA) use social intelligence to solve problems in optimization. The CA is a class of evolutionary computational models inspired from observing the cultural evolutionary process in nature. Cultural algorithms employ a basic set of knowledge sources, each related to knowledge observed in various animal species. Knowledge from these sources is then combined to influence the decisions of the individual agents in solving problems. Classification using “IF-THEN” rules comes under descriptive knowledge discovery in data mining and is the most sought out by users since they represent highly comprehensible form of knowledge. The rules have certain properties which make them useful forms of actionable knowledge to the users. The rules are evaluated using these properties represented as objective and subjective measures. The rule properties may be conflicting. Hence discovery of rules with specific properties is considered as a multi-objective optimization problem. In the current study an extended cultural algorithm which applies social intelligence in the data mining domain to present users with a set of rules optimized according to user specified metrics is proposed. Preliminary experimental results using benchmark data sets reveal that the algorithm is promising in producing rules with specific properties.