كنفرانس بين المللی مهندسی صنايع و سيستم ها

صفحه اصلی / دوازدهمین كنفرانس بين المللی مهندسی صنايع و سيستم ها

TRFMD: A Time-Decay and Diversity-Aware RFM Model for Profit-Oriented Fuzzy Customer Segmentation

نویسندگان :

Mansoureh Naderipour¹ MohammadSaeed Balouchi Javaran²

1- دانشگاه شهید باهنر کرمان 2- شهید باهنر کرمان

کلمات کلیدی :

TRFMD،customer segmentation،fuzzy C-means،time-decayed monetary value،product diversity،customer profitability،retail analytics

چکیده :

Customer segmentation is a fundamental component of retail customer relationship management; however, conventional RFM analysis assigns equal importance to historical transactions, overlooks customers’ product-purchasing breadth, and commonly relies on hard clustering methods that impose rigid segment boundaries. This study proposes TRFMD, a time-decay and diversity-aware extension of the RFM framework for profit-oriented fuzzy customer segmentation. The proposed model integrates time-decayed monetary value with recency, frequency, conventional monetary value, and product diversity. Fuzzy C-means clustering is employed to estimate graded segment memberships and identify customers whose behavioral profiles lie between managerial segments. The model is evaluated using the UCI Online Retail II dataset, comprising 805,549 cleaned transaction lines associated with 5,878 customers. Customer-level features are constructed over a calibration period, while future revenue and repurchase behavior are assessed during a subsequent holdout period. Comparative experiments involving RFM, RFM-D, and TRFMD under both K-means and fuzzy C-means demonstrate that TRFMD provides substantially stronger future-revenue discrimination. Specifically, TRFMD-FCM achieves a holdout revenue eta-squared of 0.2133, compared with 0.0649 for conventional RFM/K-means. The resulting fuzzy segments also reveal boundary customers that hard clustering conceals. These findings indicate that TRFMD offers an interpretable and actionable framework for customer retention, reactivation, cross-selling, and marketing resource allocation.