LACRM: LLM-Guided Mining of Directed Advertiser Competition Relationships in E-Commerce Advertising
LACRM proposes a framework to model competition among e-commerce advertisers by creating directed competition graphs with continuous edge intensities. It uses a three-stage pipeline—Dual-Profile Generation, Multi-Route Retrieval, and Dynamic Behavioral Alignment—to learn how advertisers compete. The method successfully identifies latent rivals beyond historical data and shows measurable revenue gains in online bidding experiments.