The explosive growth of network information system makes the recommendation become a research hotspot. Existing recommendation systems used in the actual operation have respective defects. In web2.0 environment, tags, the item rating and the time of the user tagging the item contain important information suggesting the user's preference. This information is useful to improve the accuracy of the recommendation system. Drawing on collaborative filtering method, we suggest an item recommendation model which is considering tags, the item rating and timeliness of user's preference all together and discuss the architecture and prospect of this method.