Comparison and analysis of applications of ID3, CART decision tree models and neural network model in medical diagnosis and prognosis evaluation

Objective: To analyze the performance of each algorithm model under different processing conditions such as data preprocessing (standardization, normalization and regularization), balancing and shuffling based on the data attributes of three common research types in clinical studies as the research examples. To compare and analyze advantages and disadvantages of the decision tree model and the neural network model in clinical studies as well as their scope of application. Methods: Python was used to construct ID3 and CART decision tree models. Three typical clinical research data sets were downloaded from UCI and used to perform data preprocessing, balancing, and shuffling on the models. The model evaluation indexes included time complexity, accuracy, precision, recall and F1-Score. As for visualization, the model results, confusion matrix and ROC curve were drawn. The importance rankings of different data set attributes on the model results were also analyzed. In addition, one typical data set was selected to conduct the comparative analysis by using the neural network model. SPSS was used to perform the significance analysis of different data processing schemes. The SPSS platform was used to conduct the statistical test of the results.