Cancer has become a non-negligible problem that threatens human health in today's society. The traditional methods of cancer diagnosis usually use cell morphology, histopathology and other methods. Nowadays, the use of machine learning technology to predict cancer has become a new actionable way. This paper proposes the use of machine learning algorithms to assist breast cancer diagnosis, using a variety of algorithms such as LightGBM, Random Forests (RF), Support Vector Machines (SVM), Linear SVM, K-Nearest Neighbor (KNN), and combined with grid search algorithms, respectively constructed intelligent predictive and diagnostic models for malignant breast cancer. Finally, using the breast cancer data set of the University of Wisconsin (WCBD) hospital to conduct experiments, a classification model based on LightGBM-Gridsearchcv is proposed. Compared with the models, the LightGBM-Gridsearchcv model has a recognition accuracy of 95.9% for malignant cancer cases. The machine learning method has put forward a new research idea and method for the diagnosis of breast cancer, and provided a research direction for the promotion of intelligent medical treatment, which has very important practical significance and application value.