基于卷积神经网络的糖尿病预测

Prediction of Diabetes Based on Convolutional Neural Network

摘要: 生物技术和医学的显著进步导致了生物医学数据的大量产生。糖尿病(Diabetes mellitus, DM)作为一种常见的慢性病，在诊断和治疗过程中也产生了大量的医学数据。因此，医学数据的探索成为了一个热点。本研究旨在探讨出院病人的短期入院率。根据30天内再次入院的概率，我们可以判断这种治疗的效果。从而协助医生为患者提供更有效的治疗，以便提高患者的生活质量。在本研究中，对数据进行了处理，利用改进的卷积神经网络算法(Convolutional neural network algorithm, CNN-EI)对糖尿病病例数据集进行数据挖掘。实验结果表明，改进后的算法能够很好地处理高维、大样本的医学数据。将该方法的结果与其他先进方法的结果进行比较，其准确率为83.7%。

Abstract: The remarkable progress of biotechnology and medical science has led to the significant production of biomedical data. For diabetes mellitus (DM), a common chronic disease, a large number of medi-cal data has also been generated in the process of diagnosis and treatment. So the exploration of medical data has become a hotspot. The purpose of this study was to explore the short-term admission probability of discharged patients. According to the probability of re-admission within 30 days, we can judge the effect of this treatment, thus assisting doctors to provide more efficient treatment for patients, so as to improve the quality of life of patients. In this study, the data has been processed, and the improved convolutional neural network algorithm (CNN-EI) was used to perform data mining on the dataset of diabetic case data. The experimental results show that the improved algorithm can well perform high-dimensional and large-sample medical data. Accuracy is 83.7%. The result is compared to the result of other state-of-art methods.