Pattern Drift Detection in Association Analysis Based on Relation Entropy and J Measure
Yang Ying-ji · 2016
Aiming at the problem that traditional concept drift detection technology cannot be applied in association analysis directly,this paper proposed a method of concept drift detection based on relation entropy and J measure.Four features such as relation count,relation entropy,windows count and J measure are extracted.Concept drift is detected and located through hypothesis test of features distribution in two sliding windows.Event set and event pair set mostly effected by concept drift which can provide support to adjustment and refreshment of concept are obtained.Experimental results show that this method is accurate and feasible.