Using a Partition System to Improve the Performance of the Apriori Algorithm in Speeding Up Itemset Frequency Search Process

The apriori algorithm uses minimum support and minimum confidence to determine appropriate itemset rules for decision making. The problem faced in this research is how to improve the performance of the a priori algorithm in the process of searching for itemset frequencies using data partition techniques, and be able to produce optimal and consistent rules. To overcome this problem, the author implemented the a priori method and partition system to improve the performance of the a priori algorithm for the itemset frequency search process by taking public data in the form of supermarket transaction data. In this research, the performance of the a priori algorithm was tested with and without a partition system. The data used in this research consists of 350 transaction data from 1784 records with a 4-itemset pattern, minimum support value of 20% and minimum confidence of 0.5 with the best standard rules for determining minimum confidence of 0.8. Based on this research carried out, the research results obtained are that for comparison of time and memory usage the apriori algorithm with a partition system is much faster than the apriori algorithm without a partition system, while memory usage is relatively less for the apriori algorithm with the system than the apriori algorithm without a partition system.