This paper analyzes electronic commerce trading patterns evolution from a perspective of Agent technology and electronic commerce applications. According to characteristic of customer and business, it raises Model of Multi-Agent based on personalized transactions in Electronic Commerce which can realize personalized trusteeship without participation of customer. And this paper examines about role of agent connected with electronic commerce in present web and agents for electronic commerce in semantic web which is observed by the next generation web, proposes electronic commerce platform that apply semantic web technology, we want to use Bayesian Networks to find the potential rules through examples of all transactions (ex. buying behavior) for businesses.