Spare Parts Inventory Management Literature and Direction Towards the Use of Data Mining Technique: A Review

2. Data Mining Studies Related To Item Dependency

A considerable number of researchers studied the dependency of items using data mining. Data mining application has received a lot of attention in the last decade from various sectors like medical service, intrusion detection, manufacturing and customer relationship management etc. Various methods of data mining such as itemset mining, mining sequential pattern, association rule, classification, prediction, clustering, regression etc. are applied for the above areas (Han, J. and Kamber, 2006). Itemset mining is the first fundamental task for further data mining methods like association rule discovery, sequential pattern mining, classification and prediction etc. An itemset is called a frequent item set when support (frequency of occurrence in database) value of itemset is greater than or equal to minimum or threshold support value which is usually set by decision makers. In frequent item set mining area, various authors proposed their algorithms related to candidate generation and reduction of execution time. The popular approaches are Apriori and subsequent apriori-like algorithms (Agrawal et al., 1993) and pattern growth methods (Agarwal et al. 2001). These approaches assume all items as binary variables considering if they are consumed or bought or not. Hence few researchers proposed extension of itemset mining called itemset mining with quantities considering the consumption of item along with quantity (Srikant et al., 1996). Another important study is weighted itemset mining which considers the occurrence of item along with its weight or importance (Yun et al., 2005) which was not covered in the earlier studies. The acceptances of all these frequent itemset mining methods are selected based on the threshold support or minimum frequency values.

The Association rule mining technique discovers the relationships among items present in transactions (Agrawal et al., 1993). In fact, data mining application has very less contribution towards the inventory management and a brief review of literature on data mining application for inventory management is provided here. Brijs et al. (1999) used association rules for selecting items considering relationships among retail items by discovering frequent itemsets and found out the profitability per set of items by identifying the cross-sales effect of product items and using this information for better product selection. This model was extended by Brijs et al. (2000) to enable retailers to add category restrictions. Similarly, an extensive study was performed by Brijs (2002) on retail market basket analysis using quantitative modeling approach. Wong et al. (2005) followed the Brijs’s work and proposed a methodology for suggesting recommendations from analysis of itemsets and the application of the concepts of association rules and selection of maximal profit item are investigated considering cross-selling effect.