| Abstrak/Abstract |
Traditional association rule mining algorithm only generates a set of rules from frequent itemset, the rules obtained cannot generate rules from high-utility itemset. This is because the framework that’s being used to obtain rules from traditional association rule is support-confidence while getting high-utility itemset association rules uses the utility-confidence framework. The model for high-utility association rule mining proposed is using particle swarm optimization. The fitness function to get high-utility association rules does not use support-confidence but uses the utility-confidence framework. The association rule mining model of high-utility itemset does not look for high-utility itemset first but together with the high-utility itemset mining process. The high-utility association rule mining using the particle swarm optimization approach has better rule set quality than using the traditional approach, Apriori. Testing with five datasets: chess, connect, mushroom, accident, and foodmart, shows the average utility-confidence obtained using our proposed method is above 88%. |
