Performance Improvement of Apriori Algorithm with Transaction Reduction and Hash Based Approach

The apriori algorithm is an algorithm that generates candidate itemsets incrementally and recursively to calculate and combine itemsets until no candidate itemset. Apriori algorithm has limitations, namely requiring a lot of time to scan the database to calculate support and confidence. In this study, the performance improvement is the execution time parameter for data scanning, and then researchers propose a method to speed up the performance of the apriori algorithm by combining the transaction reduction and hash-based methods, where the transaction reduction method is expected to be able to reduce the data to be scanned in the database by eliminating data whose value is below the minimum support. Then the hashbased method is able to reduce the data to be scanned in the database because the data scanned is only the hash code, while the large itemsets data will be represented by the hash code without making the performance of the validity of the rules formed does not decrease. Because the most useful information from finding patterns using the apriori algorithm is the rules formed. So the time required to execute a large database can be reduced a lot with association rules that run according to the procedure. From the results of tests conducted by researchers, at 2-itemsets in the second iteration the processing time is reduced about 20 times faster, then in the 3-itemsets in the third iteration the processing time was reduced by around 790 times faster compared to the apriori algorithm without transaction reduction and hash based methods.