RFM analysis for customer segmentation using feature generation and accelerated deep learning for prediction

In this project, we propose a feature generation method for customer segmentation using an online retail store dataset. Also, we perform a comparative cluster analysis between the proposed method and the traditional method with K-means. We show that among the top-7 classifiers reported in previous studies, Multi-layer Perceptron (MLP) is the best. Then, we use the Genetic Algorithm (GA) to improve the speed and accuracy of MLP during the prediction phase.

本研究基于在线零售商店数据集，提出一种用于客户细分的特征生成方法。同时，针对所提方法与基于K均值聚类（K-means）的传统聚类方法，开展对比聚类分析。本研究证实，在既往研究报道的前7种分类器中，多层感知机（Multi-layer Perceptron，MLP）性能最优。随后，本研究采用遗传算法（Genetic Algorithm，GA）对预测阶段的MLP模型进行优化，以提升其运算速度与预测精度。