After the destructive earthquake, it is urgent that the damaged houses, infrastructure and industrial projects and other related equipment and facilities are to be rehabilitated and restored. How to carry out reconstruction investment plans of various sectors became a major concern to the government. To meet the needs of rapid recovery of production and life in affected areas, the method based on BP neural network is proposed to estimate the post-earthquake rehabilitation cost of industrial, agricultural and service projects. That method considers the macro-economic characteristics of every damaged area. It can provide some guidance for the government that need to make decisions of post-earthquake recovery and reconstruction plans.