The invention discloses a transformer fuzzy prudent reasoning fault diagnosis method based on cost sensitive learning, and belongs to the field of transformer state evaluation and fault diagnosis. According to the invention, on the basis of acquiring a transformer state evaluation initial sample set and setting a cost sensitive initial matrix, firstly, by combining a Sigmoid multiattribute softening decision, a transformer fault diagnosis multi-class support vector matrix model is constructed; then, a support vector matrix is subjected to normalized ordered weighted averaging, and a fuzzy prudent membership degree weight is calculated; and finally, complementary confidence allocation and information fusion based on a PCR5 method are carried out on a weighed fuzzy prudent membership degree,and based on a confidence allocation fusion final value, transformer fault diagnosis decision determination is carried out. In such process, minimization of an error diagnosis sample number is used as an optimal object, a cost punishment element corresponding the cost sensitive matrix is iteratively corrected on the basis of the fuzzy prudent evidence reasoning process, and an online learning function of the fault diagnosis model is achieved.