A novel approach for unsupervised learning of transaction data

Abstract

Incremental clustering is a technique which can be applied when the dataset is not constant and keeps updating. Normally when kmeans clustering is applied and if the dataset is modified then the clustering must be done from start. Similarly, for maximum capture procedure proposed in our previous research the clustering task must be carried from the start. In this paper, we propose an incremental approach for clustering transaction data which can be used for customer segmentation and other related applications. Experiments are conducted and three approaches are compared in terms of CPU utilization. It is observed that incremental approach required less CPU utilization.