Deep Mining Learning and Application of Health Data

The information in the medical database is extremely rich, which may contain medical images of patients, relevant pathological parameters, laboratory and measurement results, diagnostic records, and related parameter data. In the actual clinical diagnosis process, by analyzing the basic information provided by patients and inputting it into the diagnosis system, a more accurate diagnosis conclusion can be given. In this paper, DL (Deep Learning) technology is used to establish a learning model for deep mining of medical health data, and it is applied to mining the relationship between medical health data and diabetes screening, and a simple method for diabetes screening is realized. Firstly, the diabetes data set is normalized to meet the input requirements of CNN (convolutional neural network). Because of the structural characteristics of diabetes diagnosis data, the typical CNN was improved, the model parameters were determined through unsupervised and supervised training, and the model was established. The research results show that AUC is 0.767, which means that the improved CNN is very close to the ideal situation and can achieve good diagnosis results. Compared with other classifiers, the experimental results show that the improved CNN algorithm has good classification accuracy and stability, and the accuracy, recall, and F1 score are 0.89, 0.898, and 0.859, respectively. Therefore, the good accuracy and algorithm performance of the model provide the possibility for the application of the algorithm in practice.