With the prosperity of new forms of media, the social environment of universities is increasingly complex. The change not only brings development opportunities for university, but also makes a number of uncertain factors increase rapidly, so that colleges and universities have to face more and more frequent reputation crisis. To reduce the harm of crisis, it is particularly important to focus on pre-crisis warning system construction. Colleges and universities should spend a small amount of money on prevention, rather than on treatment. While the existing literature on crisis early warning mechanism does not discuss deeply, this paper adopts the BP neural network method innovatively and uses new techniques such as data mining to research crisis early warning mechanism of Reputation of colleges and universities in the new media environment. Besides, this paper take subordinates MIIT universities as an example to make the analysis of the current construction of crisis early warning mechanism of reputation of colleges and universities status quo, so that it can make appropriate recommendations for improvement.