ANALYSIS OF CUSTOMER PURCHASE DATA USING A PROPORTIONAL HAZARD MIXTURE MODEL

DOI Web Site

Bibliographic Information

Other Title
  • 比例ハザード性を仮定した混合分布による購買履歴データの分析
  • ヒレイ ハザードセイ オ カテイ シタ コンゴウ ブンプ ニ ヨル コウバイ リレキ データ ノ ブンセキ

Search this article

Description

Retail discounts are widely implemented at stores and e-commerce sites (EC sites). We propose a new mixture distribution model for classifying customers based on their reactions to discounts. For a mixture distribution with an ordered structure, a proportional hazard model is a natural fit. Using a bootstrap method, we evaluate the estimators obtained by an expectation-maximization (EM) algorithm. The usefulness of the proposed model is tested by analyzing the EC sites' purchase history data. Evaluation using recency/frequency/monetary (RFM) analysis reveals a cluster of good customers characterized by a high probability of buying discounted items.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top