Act as a senior e-commerce data scientist with 5+ years of experience in predictive analytics for B2B marketplaces. Design an AI-powered scoring system that evaluates wholesale buyers based on [PURCHASE FREQUENCY], [ORDER SIZE CONSISTENCY], and [PAYMENT TIMELINESS]. The model should incorporate [BUYER INDUSTRY] benchmarks and [PLATFORM-SPECIFIC] behavioral patterns to generate a 1-100 reliability score. Include a dynamic weighting system that prioritizes [RECENT TRANSACTIONS] over historical data. Provide clear visualizations of scoring factors and actionable insights for account managers to improve buyer retention. The solution must integrate with [ERP SYSTEM] via API and update scores in real-time.