High numeric coherent association rule mining with a particular categorical consequent class attribute

It has been observed that sometimes in Numeric Association Rule Mining (NARM), it is important to understand the association between a numeric attribute and a specific categorical consequent class attribute. NARM divides the domain of numeric attributes sub-domains without considering particular categorical consequent class attribute. Apart from this, it may also suffer from support-confidence conflict problem. If the domain of attributes is divided into large sub-domains, some rules may generate extremely low confidence. Thus to solve this problem, this research work proposes a bipartition technique which takes minimization of XOR-true value into consideration and produces high confidence and reliable numeric coherent rules. We proposed a High Numeric Coherent Association Rule Mining (HNCARM) algorithm which contains two steps - a pre-processing and post processing step. In preprocessing step, numeric attributes are converted into Boolean attributes and in post processing step, rules, having particular categorical class attribute in its consequent, are extracted. The proposed methodology has been implemented with two benchmark datasets and generates encouraging results with strong and efficient numeric coherent rules.