基于c-均值聚类的粗糙集神经网络在心脏病诊断中的应用

The Application of Rough Set-Neural Network Based on c-Means Clustering in Heart Disease Diagnosation

采用c-均值聚类法将决策表中的连续条件属性进行离散化,用粗糙集处理离散化后的决策表系统得到简化规则,然后将规则集输入BP神经网络进行训练,并对测试集进行预测.以此模型对一组有关心脏病诊断的数据进行处理,得到的预测判准率达85%,而单独使用粗糙集或BPNN进行预测,则判准率分别为76%和82%;若在粗糙集和BPNN联用模型中,对原始数据采用传统的等距离离散化和等频率离散化等离散化方法,预测判准率则分别只有53%和77%.

By using c-means clustering, an algorithm is presented to discretize condition attributes in a decision table. The simplified rules are available with the rough set theory processing the discretized decision table system, and the data set obtained is then introduced to the BP network for training and testing. Based on the present model a group of data form heart diseases is processed, achieving a testing correctness percentage of 85 while the other two are 76% and 82% respectively. In the combined model of rough set and BP neural network, the tradition algorithm, i.e. the equation-distance and the equation-frequency for discretization can only have 53% and 77% of testing correctness.