Elastin-like polypeptides (ELPs) have been widely used to promote the development of a variety of smart biomaterials. Transition temperature is a key attribute of ELPs central to ELPs researches. Therefore, it is essential to establish predictive models of transition temperature that are both computationally efficient and reliable by employing simple parameters. Back propagation neural network (BPNN), a powerful feed-forward algorithm designed to solve problems with overwhelming complexity, has been successfully used in non-linear predictor model. In this study, two pH-sensitive ELPs were expressed, purified and determined for temperature transition across a range of pH. The pH, concentration and molecular weight (MW) as well as isoelectric point (PI) and pseudo amino acid (PseAA) of these two ELPs were adopted as input parameters. Support vector machine (SVM) and back propagation neural network (BPNN) were performed respectively. We selected Lamda (λ) value by training set and evaluated predictor model by jackknife test that combined with Uniform Design (UD). According to the results of BPNN and SVM, whose mean absolute error (MAE) of BPNN model jackknife test were 4.80 and 4.95 respectively, the predictive ability of BPNN is a minor improvement over SVM. Applying Mackay's data, MAE of BPNN jackknife test was 2.02, while the MAE between experimental and predicted transition temperature was 2.30 in Mackay's predictor model. Compared with Mackay predictor model, the enhancement in the accuracy indicates that the proposed BPNN method could play a complementary role to predict ELPs transition temperature.