The application of improved BP neural network in the diagnosis of breast tumors

The traditional BP neural network is improved and developed in this paper. When the iteration of Levenberg-Marquardt takes the place of Gradient descent algorithm, the network convergence rate is improved greatly. After the analysis of breast tumors offered by Dr. William H. Wolberg from University of Wisconsin Hospitals, 9 parameters reflecting the characteristics of breast tumors are concluded. Based on the improved BP neural network, the simulation model of breast tumors is founded. Within the 83 groups of testing data, benign diagnosis rate is 100%, while malign diagnosis rate is 96.6%.