The prediction and evaluation of mineral resources is a systematic process, which is composed of multiple stages including data collection, data collation, information extraction, information synthesis and utilization, and prediction and evaluation of mineral resources from the information integration, thus to the ultimate results. It is because of the complexity and diverse changes of geology and the incompleteness of human knowledge and backward cognition that makes the prediction and evaluation of mineral resources with large uncertainties in each stage. And these uncertainties is very likely to result from the former period. All these lead to inaccuraries and errors in mineral resource prediction, and also cause quite a lot of uncertainty accumulation and propagation. At present, with more and more difficulties to find mineral resources and the increasing economic risks of mining industries, we have the urgent need for a better understanding of uncertainies in mineral prediction and evaluation. This research focuses on the the sources of uncertainties in the mineral resources prediction to offer the propagation model and algorithm accordingly.