基于PCA和BP神经网络的糖尿病预测

Diabetes Prediction Based on PCA and BP Neural Network

摘要: 糖尿病的发病率正在逐年上升且向低龄化发展，对我国乃至世界的健康安全造成了严重的影响，因此有必要对糖尿病的预测进行研究。本文对皮马印第安人糖尿病数据集进行分类，首先使用主成分分析法将数据从8维降到了3维，接着使用这3维数据建立BP神经网络模型。将基于PCA和BP神经网络的模型与单纯的BP神经网络模型进行对比。结果表明，基于PCA和BP神经网络的模型在精准率、召回率、F值、查准率和Matthews相关系数MCC 5项性能指标上均明显优于BP神经网络，可以作为糖尿病预测的一种有效方法。

Abstract: The incidence of diabetes mellitus is increasing year by year and towards lower age, which has a serious impact on the health and safety of our country and the world, so there is a need to study the prediction of diabetes mellitus. In this paper, the Pima Indians diabetes dataset was categorized by first reducing the data from 8 dimensions to 3 dimensions using Principal Component Analysis (PCA), followed by BP neural network modeling using these 3 dimensions of data. The PCA and BP neural network based model was compared with the BP neural network model alone. The results show that the model based on PCA and BP neural network is significantly better than BP neural network in five performance indicators: precision, recall, F-value, checking accuracy and Matthews correlation coefficient MCC, and it can be used as an effective method for diabetes prediction.