Customer Profitability Prediction In Online Retail Industry

Abstract

This study investigates the prediction of customer profitability in business context with the aim of better interpreting the customer purchasing behavior and adopting proper selling strategy. In this paper, we also extend the existing approach to predict this metric. The time series of customer group of low to high level of profitability was generated by the Recency, Frequency, and Monetary (RFM) model and k-means clustering algorithm from transaction data. A new measure that is Clumpiness factor was also explored to increase prediction accuracy. Finally, different models of multilayer perceptron neural network were trained and utilized to evaluate the model performance. The experimental results have shown a promising predictability of the RFM as well as clumpiness element.