Comparison of Apriori vs. FP-Growth Algorithm with Partitioned Data, A Case Study: Recommendations for Selection of Specialization Courses Ianatush Sholihah, Rosihan Ari Yuana, Nurcahya Pradana Taufik Prakisya
Informatics and Computer Engineering Education, Faculty of Teacher Training and Education, Sebelas Maret University Indonesia
Abstract
Association Rule Mining (ARM) is a research topic in data mining that aims to find interesting correlations or patterns in databases. The ARM method has two algorithms that are populer in searching for item patterns in a transaction data, namely Apriori and FP-Growth. The Apriori algorithm is often used because the process is simple but takes a long time to process if the data used is large enough. The FP-Growth algorithm is here to provide an alternative way to overcome the shortcomings of the Apriori algorithm. But in some cases, FP-Growth takes longer than Apriori. To overcome this problem the data to be processed is partitioned first. A comparative analysis of the Apriori and FP-Growth algorithms was carried out by partitioning processed data in case studies of recommendation specialization courses using the Knowledge Discovery in Database (KDD) method. The analysis is carried out based on the length of execution time, memory usage, and the Lift Ratio value of each algorithm. Based on the test results, the best algorithm for recommendation for specialization courses is the FP-Growth algorithm because the required execution time is 25.4% faster. The FP-Growth algorithm is also able to produce more rules with 3.14% less memory consumption and produces a higher support, confidence, and lift values than the Apriori algorithm.
Keywords: Apriori, Association Rule, Data Mining, FP-Growth, Study Plan