The invention relates to a network-level bridge structure performance evaluation and prediction method, which comprises the following steps: S1, collecting a detection report of each bridge over the years, extracting technical condition scores, bridge ages, structure types, traffic volumes and maintenance behavior information of each year, and constructing a relational database; s2, based on the data in the relational database, training and checking the established structural performance degradation model; s3, predicting the overall structure of the bridge in the regional road network and theperformance change trend of the local component by using the trained structural performance degradation model; and S4, giving an optimized overhaul scheme and a maintenance strategy of bridges in theregional road network. Compared with the prior art, the method has the advantages that the data mining technology is integrated, mass detection data accumulated under the condition of long-term bridgeinspection work is effectively and fully utilized, the neural network model is established, the extracted data is converted into valuable knowledge in the bridge management and maintenance field, andthe network-level bridge structure performance evaluation prediction and decision support are realized.