Scoring recommendations and explanations with a probabilistic user model

摘要: A data processing system generates recommendations for on-line shopping by scoring matching the customer's cart contents using assessing and ranking each candidate recommendation expected incremental margin associated with being issued (as compared to not issued) taking into consideration historical associations, knowledge of layout site, complexity product sold, user's session behavior, quality selling point messages, life cycle, substitutability, demographics and/or other considerations relating customer purchase environment. In an illustrative implementation, inputs (such as relevance, exposure, clarity pitch strength) are included in a probabilistic framework Bayesian network) score effectiveness messages comparing outcome (e.g., likelihood or resulting from given recommendation) against non-recommendation if no is issued). addition, may also be used select message inclusion selected relative strength factoring user profile match factor that matches various case profiles).