To make full use of maintainability multi-source prior data and improve the fusion accuracy of prior distributions to ensure the accuracy of maintainability verification results, a prior distribution determination method for maintainability multi-source conflict evidence data fusion was proposed to solve the problem of conflicts in multi-source data. Fully mining and extracting the feature information of multi-source data, constructing evidence mass function based on sample size, distribution characteristics and data importance respectively, comprehensively considering the correlation between evidence and the uncertainty of evidence itself, introducing angle cosine to measure the degree of conflict between evidence. By combining the support degree and uncertainty of the evidence, the multi-source conflict evidence data fusion model was established to achieve the effective fusion of multi-source data and determine the comprehensive prior distribution. Combined with the analysis of two cases, the proposed method is proved to be effective and feasible.