Identification of CTG Based on BP Neural Network Optimized by PSO

Aiming at the normal, atypical and abnormal cardiotocography (CTG), the BP neural network (BPNN) classification model was created. The particle swarm optimization (PSO) technique was used to optimize the initial weights and threshold value of the neural network based on the inherent local minimum problem of BPNN and the good global convergence and global search ability of PSO. The CTG set of the UCI was used to test the algorithm. The simulative results indicated that the recognition accuracy of the normal, atypical and abnormal CTG reached 98.35%", "95.48% and 98.46% based on PSO-BP, more than GA-BP and traditional BP. The model has good ability both in learning and generalization and the algorithm can also effectively be used in other signal processing fields.