基于特征选择和概率神经网络的心脏病预测

Heart disease prediction based on feature selection approach and probabilistic neural network

面对我国心血管疾病患病人数的不断增长,针对心血管疾病的预测,利用监护系统获得医疗数据,寻找出合适的疾病预测方法,及时发现并解决健康问题,创新性地提出一种基于概率神经网络和遗传算法的心脏病预测模型。其中,使用概率神经网络作为分类器,遗传算法进行特征选择。模型分为三个阶段:首先,采用标准的UCI数据库中心脏病数据集进行预处理;然后,提供一种基于遗传算法的包裹式特征选择方法来选择显著特征;最后,使用概率神经网络训练得到预测模型。实验结果表明,相较于其他模式识别方法,提出的模型使用更少的特征取得了更高的准确率。通过特征选择算法可以获得显著特征,利用多种机器学习算法在经过特征选择后形成的新数据集上训练,精度也普遍得到提升。

With the increase of patients with cardiovascular diseases in China, the monitoring system is used to obtain medical data to find out the appropriate disease prediction methods for the prediction of cardiovascular diseases, so as to identify and solve health problems timely. Therefore,a heart disease prediction model based on probabilistic neural network(PNN)and genetic algorithm(GA) is proposed innovatively. In the model,the PNN is used as the classifier,and the GA is used for feature selection. The establishment of the prediction model is divided into three phases:the heart disease data set in the standard UCI database is used for preprocessing,a wrapping type of feature selection method based on GA is adopted to select the most significant feature,and the PNN is used to train the prediction model. The experimental results show that,in comparison with the other pattern recognition methods,the proposed model can achieve higher accuracy rate with fewer features.Significant features can be obtained by feature selection algorithm. Various machine learning algorithms are trained on the new data set formed after feature selection,and the accuracy of each trained algorithm is generally improved.