Research and development activities (R&D) as the core competitiveness in the high-tech enterprises play an extremely important role and far-reaching significance. If the high-tech enterprises want to realize the effective management to the R&D activities, and improve their success rate and economic benefits, it is necessary to conduct a scientific evaluation for the R&D performance. To measure the R&D performance of the high-tech enterprises scientifically and accurately, this paper overcomes the shortcoming of tradition linear R&D measuring method, proposes a measuring method which establishes a performance evaluation system combined with Kirkpatrick model and BP neural network. The performance evaluation of 16 enterprises in Hebei Province shows that the results given by this model are reliable, and this method to evaluate the R&D performance is feasible.