Article Preview
Top1. Introduction
The knowledge discovery from association rule mining represents the mined rules from transactional databases with the description of frequent itemsets that can be useful in major applications, include, e-commerce (OzgurCakir, et.al,2012), web recommendation (J. Bhavithra, et. al., 2018), decision making in expert system (Philippe Fournier, et, al., 2017) and healthcare. Association rule mining (ARM) depends on traditional support-confidence framework and it takes the uniform importance of items when generating frequent itemsets i.e. it is not considered either profit or utility of items during frequent itemsets generation. Aim of ARM is to discover the frequent itemsets based on satisfaction of minimum support confidence. The support (Jiawei Han, et.al) of association rule A→B is defined as percentage of transactions containing the itemset. Strong association rule satisfies the minimum confidence, where confidence is defined as percentage of transactions contain the itemset B (for the rule A→B) among the set of transactions containing the A. The problem of ARM is that it considers only frequency of items over to set of transaction and it may not consider the associated profit or utility of items. Both frequency occurrence and profit of items are showing the significant importance in deriving of more accurate financial and decision-making strategies in real life association mining applications.
The techniques of Apriori and FP-growth (Jiawei Han, et.al) in data mining are widely used for generation of frequent itemsets with candidate generation and without candidate generations respectively. They primarily depend on frequency count (or support count) in the process of deriving frequent itemsets, however, they are unable to consider the profit of utility values of itemsets. Other mining methods i.e.,high utility itemset (HUI) Miner (Mengchi Liu, et.al, 2012) and fastest high utility itemset miner (FHM) (Philippe Fournier, et. al, 2014) to overcome this problem by construction of utility-confidence framework instead of support confidence framework. The key objective of support-confidence framework is to derive the rules that are interested to users based on satisfaction of two thresholds i.e., minimum support and minimum confidence, however, they cannot represent semantic measure among items. Another framework of utility- confidence (K Rajendra Prasad, 2017) which exactly defines the user interested rules based on semantic measure among items with utilityvalues. The HUI Miner and FHM uses the utility-confidence framework for derivation of utility-based association rules.