Optimizing the Performance of the ECLAT Algorithm using K-means Clustering for Frequently Itemset Mining for Large Datasets

Data mining is an approach to identify the key pattern in Frequent Itemset Mining (FIM) and to find from historical data into very useful data. ECLAT is an aspect of association rule that is applied to determine the relationship connection in item supports. By intersecting the transaction id sets according to the corresponding k^th item sets based on depth-first search, ECLAT algorithm computes the support items. During searching best-selling products, ECLAT requires a lot of memory and processing time for large dataset because of the amount of transaction id sets. By merging the K-means clustering algorithm and ECLAT algorithm, the proposed method aims to determine how the K-means clustering algorithm enhances the performance of ECLAT algorithm to generate support items for large dataset. The retail data items from 100000 to 400000 UK consumers were used to retrieve the support items of the most popular selling goods. Firstly, K-means algorithm grouped the same purchased products to cluster the top selling items and then ECLAT algorithm searches the support items from clustering the famous selling products. According to the experimental results, it can be seen that the silhouette score value in this research is 0.586 to indicate a high-quality cluster, the ideal value of k is 8 clusters between 100,000 and 400,000 data items. ECLAT can search the support items based on the clustering of the top eight selling supplies by setting a minimum support value of 0.00001. The proposed method reduces memory usage and execution time to retrieve the support elements for most popular selling items than the ECLAT algorithm for massive dataset.