Research on Intelligent Diagnosis Algorithm of Diseases Based on Machine Learning

When the positive and negative sample attribute characteristics of data are similar, most machine learning algorithms often fail to classify the data correctly in the diagnosis of a single disease. In this paper, we propose a confidence interval based neural network algorithm, which can overcome the above shortcomings very well. Firstly, the proposed algorithm normalizes the training samples. Secondly, it uses Principal Component Analysis (PCA) to reduce the dimension to obtain the optimal features, and then uses the Back Propagation (BP) neural network to train the features to obtain the diagnostic model. Before using the diagnostic model to predict the samples, we firstly obtains a confidence interval for the distribution of the output values of the training samples, and then determines the final prediction result according to the confidence interval in which the predicted values fall. We run the proposed algorithm based on the disease data set in University of California Irvine (UCI) database to carry out single disease intelligent diagnosis, and compare it with other classical machine learning algorithms such as decision tree, support vector machine, random forest, K-nearest neighbor and so on. The results show that predicting accuracy of the proposed algorithm is significantly better than that of other classical algorithms mentioned above.