An Efficient Multi-set HPID3 Algorithm based on RFM Model

Abstract

Data mining is a latest emerging technique, which is mainly used to inspect large database in order to discover hidden knowledge and information about customers' behaviors. With the increasing contest in the retail industry, the main focus of superstore is to classify valuable customers accurately and quickly among the large volume of data. The decision tree algorithm is a more general data classification function algorithm based on machine learning. In this paper the concept of Recency, Frequency and Monetary is introduced, which is usually used by marketing investigators to develop marketing strategies, to find important patterns. Conventional ID3 algorithm is modified by horizontally splitting the sample of customer purchasing RFM dataset and then classification rules are discovered to predict future customer behaviors by matching pattern. The dataset has been accessed from blood transfusion service center and has 5 attributes and 748 instances. The experimental result shows that the proposed HPID3 is more effective than conventional ID3 in terms of accuracy and processing speed.