optical breast spectroscopy (OBS) can predict breast density and can be used for risk assessment of breast cancer (BC). We have used this method to study the classification and prediction of breast hyperplasia. In this paper, the collected spectral data of clinical breast hyperplasia and normal human breast were classified and predicted by using machine learning Bayesian model, decision tree and k-nearest neighbor model (KNN). The research shows that the accuracy of machine learning algorithm for classification and diagnosis of breast hyperplasia can reach 93%. The combination of machine learning algorithm and practical medical diagnosis technology is of great significance, and it can be further developed to improve the efficiency and accuracy of clinical diagnosis.