In order to reduce the overfitting phenomenon of standard sample and data set selection in the training process and improve the accuracy of pneumonia recognition,an algorithm for pneumonia recognition by fitting constructed convolutional neural network with model is proposed.The algorithm firstly preprocess-es the image using image threshold segmentation technique,then combines the convolutional neural network and back propagation algorithm to optimize the convolutional neural network so that the extracted features tend to be more abstract and more expressive,which can avoid the overfitting phenomenon while improving the convergence degree.Besides,the algorithm is proved by clinical data experiments that the scheme is more accurate than the existing commonly studied classification algorithms.And,clinical data shows that the scheme has improved the accuracy and recognition rate compared with the commonly studied classifica-tion algorithms,and has improved the overall interactive processing of poor data response,which has pefect the pneumonia discrimination problem to a certain extent.
维普
万方数据