METHOD AND APPARATUS FOR RECOMMENDATION MARKETING VIA DEEP REINFORCEMENT LEARNING

摘要

Provided is a method for recommendation marketing via deep reinforcement learning. A deep reinforcement learning system comprises an intelligent body and an execution environment, wherein the intelligent body is used for determining, according to a marketing policy, a marketing action for state information of the execution environment. The method comprises: acquiring, from an execution environment and with respect to multiple effect targets on a marketing effect link, multiple execution results respectively corresponding to users; according to the acquired multiple execution results, determining a bonus score of reinforcement learning; and then returning the bonus score to the intelligent body for the intelligent body to update the marketing policy thereof.