Article Preview
TopIntroduction
Analyzing purchase behavior among various consumer choices with respect to numerous products and services during a single shopping experience to realize possible correlations has been a popular method in order to take the shopping experience on a whole new level. This technique is also popular as association rule-mining due to its involvement in detecting hidden patterns from huge transactional database through the extraction of associations or co-occurrences in transactional data (Kim et al., 2012). In a study by Mostafa (2015), novel data mining technique is applied for the first time to investigate consumer behavior in Kuwait. Data mining has always played a crucial role in Customer Relationship Management helping with the retrieval of sensible information from large sets of data only to further improve the management of relationship.
A traditional problem, as argued by Burke (1996), among retailers is the determination of products’ group by brand or product type. Through the method of identifying consumer purchase behavior by obtaining linkages or co-occurrences from stores’ transaction logs Market Basket Analysis helps in strategically designing a layout such that products associated with each other are juxtaposed. Several other marketing applications like cross-selling (e.g., Russell and Petersen, 2000), designing promotion campaigns (e.g., Abraham and Lodish, 1993) are resolved with the help of Market Basket Analysis. A-priori algorithm (Agrawal et al., 1993), one of the most prevalent methods till date to obtain associations from transactions uses the concept of minimum support, assessing the frequency of both P and Q in the transactions, and minimum confidence, assessing the correctness of the rule, only to imply that a pattern of P Q means if P is bought then Q will be bought as well. However, A-priori (Agrawal et al., 1993) fails to consider the utility because of its prime concern on frequency. The term utility can be defined as usefulness of the presence of item sets in transactional logs, and is expressed as a quantity in terms of profit or sales, etc. of an item (Srivastava et al., 2018). The limitations of existing frequent item mining techniques have been overcome through the amalgamation of both high frequency and high utility while determining useful associations (Srivastava et al., 2018).
The associations discovered from a plethora of data has many applications. Recommendation Engine being another one of them. People are unaware of their own requirements until they are told so. A study found out that purchase behavior of shoppers who experience personalization, ranging from an influenced future purchase to an immediate added order, is amenable. A system to filter information aiming to analyze and estimate the ‘rating’ that user would give to an item and accordingly make recommendations is one of the ways to personalize the shopping experience. In a study by Hallowell (1996), there was a correlation between customer satisfaction and customer retention, and between customer retention and profitability. The overall effects, including the before and after effects, of a recommendation system combined with pure search (querying) and browsing (directed or non-directed), they allow users facing a huge amount of information to navigate that information in an efficient and satisfying way (Davidson et al., 2010).