Recently the Ministry of Education is actively promoting the overall implementation of passport on students' physical fitness. To meet the needs of the schools targeting developing students' physical fitness, this study attempted to apply back-propagation neural network to establish a diagnosis model of exercise strength and duration. This study took ninety college students as subjects who were aged from 19 to 22 years. Data of the participants' age, aerobic fitness、and body composition were collected. Then the three-layer structure in the backpropagation neural network-the input layer, the hidden layer, and the output layer-was programmed in C++ programming language to compute the network learning and recall processes and thus infer a diagnosis model of the exercise strength and duration fitness of the subjects. The finding of this study showed that after the computation of the network simulation, the back-propagation neural network was proven to be capable of providing a useful diagnosis model. Therefore, we consider it to be worthwhile for the appropriate authorities to develop computer programs for testing physical fitness.