Pattern of E-marketplace Customer Shopping Behavior using Tabu Search and FP-Growth Algorithm

Abstract

Pattern of customer shopping behavior can be known by analyzing market cart. This analysis is performed using Association Rule Mining (ARM) method in order to improve cross-sale. The weakness of ARM is if processed data is big data, it takes more time to process the data. To optimize the ARM, we perform merging algorithm with Improved Tabu Search (TS). The application of Improved TS algorithm as optimization algorithm for preprocessing datasets, data filtering, and sorting data closely related products on sales data can optimize the ARM processing. The method of Association Rule Mining (FP-Growth) to determine frequent K-itemset, Support value and Confidence value of data which is already sorted on TS is based on patterns which often appear in the dataset so it generates rules as reference of decision making for company. To measure the level of power of rule which has been formed, the Lift Ratio value was calculated. Based on the calculation of 97 rules produced, the lift ratio produces values > 1 of 82.54% and based on processing time, it produces the fastest data search in 1.66 seconds. When compared with previous research that uses the hybrid method, for data retrieval based on processing time, it produces the fastest data search within 12.3406 seconds, 150 seconds and 50 seconds. Previous studies have only compared the processing time of data searching without regard to validation / accuracy of data search. The test results in this study obtained more optimal results than when compared with the results of previous studies, namely in time efficiency and data mining in real time and more accurate data validation. As a conclusion, the resulting rule can be used as a reference in understanding shopping behavior patterns customer on the E-Marketplace.

ReliefF and Association Rule Mining to Determine Cervical Cancer Causes

“…AR mining can be used in different approaches for finding the relationships between features in dataset and in decisions support. The primary AR metrics that reflect benefits and strongly of rules are minimum support threshold (minsup) and minimum confidence threshold (mincon) [17]. The task for AR mining is:…”