A Human-Computer Automated Negotiation Model Based on Opponent’s Emotion and Familiarity

In e-commerce transactions, agent-based automated negotiation systems combine the advantages of artificial intelligence and bring great convenience to enterprises. However, few studies in agent-based model have considered the social characteristics of human interactions. In this paper, we propose a model with opponent’s emotion and familiarity. Also use a capsule network to predict the type of an opponent based on bidding. We use opponent type, emotion, and familiarity as inputs to the improved TD3 to predict opponent’ next round bidding. Through comparative experiments, our proposed model achieves better results in the transaction price, seller’s utility, satisfaction, negotiation process and the negotiation rounds.