In this paper, a new evaluation method for under-graduate education quality is proposed based on Artificial Intelligence Neural Network Back-Propagation (BP) algorithm and stress testing. Using this method, a publically available indicator pool is constructed, consisting of 19 variables in 4 dimensions such as Teaching Attitude, Teaching Content, Teaching Approach, and Basic Characteristic of Teachers, which impact under-graduates' mastery of knowledge and capacity building. After the BP neural network algorithm is used to learn the optimum parameters for this evaluation model, sensitivity test is applied to identify the indicators that have significant effects on the quality of education. Furthermore, scenario analysis is utilized to explore the influence of the quality of education under pre-specified situations, which provides theoretical and empirical support for evaluating under-graduate teaching, improving education quality, and enriching teacher resources.