Abstract: This research study adopted the method of user interest concept tree based on domain ontology and proposed a new multi-agent based consumer behavior forecasting model in e-Commerce to overwhelmed the limitations of outdated consumer behavior forecasting method. The algorithms consist of rough sets rule. The algorithm is used to attribute reduction for e-Commerce consumer actions prediction. With rule extraction model of rough sets, the rules of e-Commerce consumer behavior prediction are picked up. Practical example of consumer behavior prediction demonstrations that the novel proposed approach can be touched found knowledge efficiently and can be converted the obtainable rules easily. It has robust ability of fault tolerance and can recover the speed and quality of knowledge acquisition. The method has good practical value. From the test results, compared with the original method, it can effectively analyze and predict e-Commerce customers consumer behavior and can be decided that the customer’s complete ingesting trend.