Exploring Consumer Behavior: Use of Association Rules

Abstract

This paper focuses on problematic of use of association rules in exploring consumer behavior and presents selected results of applied data analyses on data collected via questionnaire survey on a sample of 1127 Czech respondents with structure close to representative sample of population the Czech Republic. The questionnaire survey deals with problematic of shopping for meat products. The objective was to explore possibilities of less frequently used data-mining techniques in processing of customer preference. For the data analyses, two methods for generating association rules are used: Apriori algorithm and FP-grow algorithm. Both of them were executed in Weka software. The Apriori algorithm seemed to be a better tool, because it has provided finer data, due to the fact that FP-growth algorithm needed reduction of preference scale to only two extreme values, because the input data must be binary. For consumer preferences we also calculated their means. This paper explores the different preferences and expectations of what customers’ favorite outlet should provide, and offer. Customers based on the type of their outlet loyalty were divided into five segments and further explored in more detail. Some of the found best association rules suggest similar patterns across the whole sample, e.g. the results suggest that the respondents for whom a quality of merchandise is a very important factor typically also base their outlet selection on freshness of products. This finding applies to all types of retail loyalty categores. Other rules seem to indicate a behavior more specific for a particular segment of customers. The results suggest that application of association rules in customer research can provide more insight and can be a good supplementary analysis for consumer data exploration when Likert scales were used.

Keywords

• knowledge discovery
• association rules
• Apriori
• consumer behavior
• marketing research
• meat products