Based On Bp Neural Network, The University Scientific Research Performance Evaluation System Research

With the coming of knowledge economy, the country has become the consensus of the whole country, the governments and universities have stepped up to the investment of science and technology. Therefore, how to evaluate outcomes of research fund investment, and how to evaluate the completion of personnel scientific research, also become a question that university research management department must be carefully consider. At present, the performance evaluation of university social science research is relatively rare. BP neural network evaluation methods have just started, there are many aspects need to be further refined and improved.In this paper, with combination of theoretical studies and empirical research, qualitative analysis and quantitative analysis of the combination of research methods, does in-depth discussion and research on the college social science research and performance evaluation. First introduced the thesis background and purpose; Secondly, on the basis of the summary and synthesis of the university research and a comprehensive concise, objective and comparable, operability setting principles to build a college social science research evaluation index system, introduce the BP neural network method. Finally, through empirical research, including the required data collection and procession, and sample learning and simulation results at the end of this study were summarized, and prospects for the future of college social science research and performance evaluation.This paper applies the Levenberg-Marquardt algorithm on the network to improve the evaluation model, and combines with the MATLAB software evaluation process. Expect of the methods and conclusions of this study provide the basis for the evaluation of university social science research performance.