Research On Intelligent Medical Diagnosis Based On Neural Network

With the rapid development of society,the rhythm of people's life changes with the changes of society.Currently,there are some keys medical problems in China,including"imbalance of medical resources","difficult to see a doctor","expensive to see a doctor","high rate of medical misdiagnosis" and so on.With the rapid development and progress of information technology,new information technologies such as artificial intelligence,cloud computing and Internet have brought new ideas to address these medical problems.Nowadays,some information technology has made breakthroughs in the medical field and achieved remarkable results.Therefore,the concept of digital medical treatment has been recognized by more and more people.Medical diagnosis is one of the core issues of digital medical treatment.Its accuracy and efficiency are of great significance.This paper mainly investigates the effects of applying several popular neural networks in medical diagnosis,including full-connected neural networks,convolutional neural networks,etc.By preprocessing multi-source medical data using some popular data technology such as Jieba,pandas and word2vec,the complex medical data are sorted out,Words are processed quantitatively,and the data are converted into one-hot binary variable or vector format,which can be recognized by computer algorithms.Then,these quantized data are fed to train three most popular deep learning algorithms,namely,fully connected neural network,convolutional neural network and word2vec+convolutional neural network.Finally,the accuracy of medical diagnosis among the three algorithms is compared with the traditional method such as the decision tree model.The best model is obtained,and a medical diagnosis service platfonn based is constructed.Quantitative data are obtained by data preprocessing.The final results show that the accuracy of Word2Vec+convolutional neural network is higher than other models,about 89%.Neural network-based intelligent medical diagnosis system is only a preliminary attempt.There are still many aspects to be improved,such as the improvement of prediction accuracy,the improvement of symptoms correlation,data preprocessing methods and so on,which will be further explored in the future.