Since the intelligence and cognition that the human beings show in the cognitive process are inevitably accompanied by uncertainty, the credibility of artificial intelligence electronic data with uncertainty and the scientific nature of evidence analysis also become a concern. In judicial practice, the credibility of artificial intelligence electronic data with uncertainty needs to be examined through distinguishing the evidence attributes between tangible electronic data and language (including natural languages and artificial languages) electronic data. At the same time, based on the Bayesian network, it is also necessary to make full use of the relationship between rule representation and concepts in three aspects of information point mining, pattern recognition and mode of evidence expression of artificial intelligence electronic data with uncertainty, to establish a cognitive model, and formally represent the regularity in electronic data information by using topological potential, so that we can ensure a scientific analysis process.