基于迭代随机森林算法的糖尿病预测

Diabetes prediction based on iterative random forest algorithm

针对印第安皮玛族成年女性糖尿病数据集进行分类。将迭代随机森林与机器学习方法如随机森林、K最近邻、基于不同核函数的支持向量机、人工神经网络、Logistic回归、梯度提升机、决策树分类等作比较。分析了分类识别精度、查准率、查全率、度量、ROC曲线和AUC值等6个指标。通过实验发现,与上述其它7种分类算法相比,针对该数据集,迭代随机森林在这些性能指标上的表现是最优的。

The diabetes data sets from adult female Pima Indians are classified first. The iterative random forest algorithm is compared with some machine learning methods such as random forest, K-nearest neighbor, support vector machine based on different kernel functions, artificial neural network, Logistic regression, gradient boosting machine, and decision tree classification. We analyze six indicators like identification accuracy,check precision, recall, F1 measurement, ROC curve and AUC value. Experiments indicate that for the classification algorithms, the iterative random forest method is optimal.