Exploratory Analysis With Association Rule Mining Algorithms in the Retail Industry

Abstract

Every year the retail sector expands quickly. These industries are becoming more competitive and difficult to operate in due to their expansion. Changing consumer buying habits, a decline in people's spending capacity and an increase in international retailers are a few of the difficulties that must be overcome. In the context of mining frequent item sets, many methods have been proposed to push various kinds of limitations inside the most well-known algorithms. This study presents an exploratory analysis for retail stores that uses market basket analysis as one of the data mining techniques to identify frequent patterns in customer purchases. The proposed method is based on comparing two algorithms: Apriori and Frequent Pattern Growth (FP- Growth). The study used a retail store dataset consisting of 522,064 rows and 7 variables. Data pre-processing was performed to clean and encode the data to be used in the model. The dataset limitation involves 25% null values in the ID column. To address this, missing customer IDs are filled with the last valid ID, assuming repeated purchases. The FP-Growth algorithm was found to be faster and more effective than the Apriori algorithm in extracting frequent item sets and generating association rules. The retail industry based on these frequent item sets is expected to increase sales by recommending highly associated items to customers.