一种基于特征组合和卷积神经网络的心脏病预测新方法

A novel method of prediction for heart disease based on convolution neural networks

针对心脏病预测难的问题,提出了一种基于特征组合和卷积神经网络的心脏病预测方法。通过特征工程对数据进行预处理,减少噪声干扰;使用特征组合算法增强样本属性关联,生成特征矩阵;设计卷积神经网络对特征矩阵进行更高级抽象。该方法在UCI Heart Disease数据集上达到了0.898 9的预测精度,优于SVM、集成学习等传统机器学习方法,可作为相关领域专家判断的重要参考。

Aiming at the difficulty in predicting heart disease, a predicting method for heart disease is proposed based on feature combinations and convolutional neural networks. This method will preprocess the data through feature engineering to reduce the noise interference. Then the feature combination algorithm is used to enhance the attribute association of the sample so as to generate the feature matrix. Finally, feature matrix is carried out more advanced abstraction by the design of convolutional neural network. This method can achieve the prediction accuracy of 0.898 9 in the dataset of UCI Heart Disease, which is superior to traditional machine learning methods such as SVM and ensemble learning, and could be used as an important reference for the judgment of domain experts.