Li Yixin，Zhang Zhongyun，He Zhiyi，Pan Zhen，Xu Xuan，Ning Jing
Chengdu Jincheng College，The National University of Malaysia，Xian International Studies University，The National University of Malaysia，The National University of Malaysia， Chengdu Jincheng College
Abstract: With the decline of mobile internet traffic dividends, issues such as low coupon redemption rates on local lifestyle service platforms (96.8% unredeemed in 2023) and resource encroachment by black markets (annual losses exceeding ¥1.8 billion) have become increasingly prominent. This study takes Platform M as the research object, leveraging 1.54 million orders and 2.82 million user behavior data points, integrating signaling theory and behavioral economics theory to construct a redemption rate prediction and precision targeting strategy system combining multimodal feature engineering and dynamic risk control. By introducing user responsiveness (Responsiveness) to extend the RFM model into an RFM-R segmentation model, the study employs a random forest algorithm (AUC=0.867) to predict redemption probabilities and integrates an isolation forest algorithm with a five-dimensional business rule matrix to identify anomalous transactions (detection rate: 72%). Empirical results show: the redemption rate of high-value sensitive users increased to 82%, while differentiated strategies boosted the overall redemption rate by 18%; dynamic risk control intercepted 12,167 anomalous orders, reducing resource waste by 18%. The study proposes a spatiotemporal coupling distribution mechanism (e.g., late-night limited-time coupons accounting for 35% of redemptions) and a senior-friendly design, optimizing user experience and complying with regulatory requirements. The outcomes provide data-driven decision support for platforms, promoting a shift toward precision and intelligent marketing, with practical implications for resource efficiency optimization, user value enhancement, and social responsibility fulfillment.
Keywords: coupon redemption rate prediction; precision targeting strategy RFM-R model; dynamic risk control mechanism; signaling theory; multimodal feature engineering; spatiotemporal coupling distribution; compliance governance.
学术联盟
