大数据分析技术领域不仅包含数据和算法,还有测试和验证等重要环节,其主要职能是对算法产生的模型(或称为学习器),就其预测的准确性、使用数据的测试集进行验证,用以评判学习器的性能。性能不好的学习器,要考虑调整参数,或者更改算法;而性能良好的学习器,其良好程度如何、是否达到满意程度等都需要相关指标精确表达。下面以 KNN 算法为例,就 Orange 平台所采用的测试验证模块的主要使用方法及指标含义详细说明。
(一)模 块[br][br] 1. 混淆矩阵模块[br][br] 混淆矩阵模块的名称是 Confusion Matrix,其模块图标是 [img]https://s21.ax1x.com/2025/02/12/pEuZH6s.png[/img],其矩阵的表达如表 2-4-3所示。[br] 表 2-4-3 Confusion Matrix 矩阵表[br][img]data:image/png;base64,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[/img][br][br] 如在汽车推荐算例中,产生的混淆矩阵模块如图 2-4-7 所示。[br][center][img]https://s21.ax1x.com/2025/02/12/pEuZxtU.png[/img][br]图 2-4-7 汽车数据案例的 KNN 混淆矩阵模块[/center][br] 图 2-4-7 显示了四个目标属性的分类情况,测试集数据共 518 个,其中沿对角线四个数据是各类的预测正确率。可以看出,unacc 类的正确率最高,达 97.7%,只有 2.3%被预测为acc。而 good 类的正确率最低,只有 29.2%,而 62.5%被预测为 acc,8.3%被预测为 unacc。可见,该算法在此场景的应用有待商榷,还需要其他算法的应用对比。[br][br] 按预测样本数量来表达,即可参见图 2-4-8。[br][center][img]https://s21.ax1x.com/2025/02/12/pEuZxtU.png[/img][br]图 2-4-8 汽车数据案例的 KNN 混淆矩阵(按样本数)[/center][br][br]2. 测试模块测试模块的名称是 Test and Score,其模块图形是[img]https://s21.ax1x.com/2025/02/12/pEuZ4k8.png[/img] 。仍然以汽车推荐算例为例,其界面如图 2-4-9 所示。[br][center][img]https://s21.ax1x.com/2025/02/12/pEueKcd.png[/img][br]图 2-4-9 汽车数据案例测试模块[/center]
(二)指 标[br][br] 从图 2-4-9 中可以看出,对测试集验证的主要指标有:AUC、CA、F1、Precision 和 Recall。在介绍这几个指标的含义及应用之前,需要介绍以下两个公式。[br][br] [math]FP_{rate}=\frac{FP}{N},[/math]指模型对负例预测的错误率,值越小越好。[br][br] [math]TP_{rate}=\frac{TP}{P},[/math] 指模型对正例预测的正确率,值越大越好。[br][br] (1)CA:Computer Accuracy 的简称,指模型的计算精度。其计算公式为[math]CA=\frac{TP+TN}{P+N},[/math]是指所有预测准确的样本量除以测试集的样本总量。[br] [br] (2)Precision:称之为精确度。其计算公式为[math]Precision=\frac{TP}{TP+FP}[/math],是指在所有预测为正例的样本中,被预测准确的样本占比多少。实际上,精确率是针对预测结果而言的,它表示的是预测为正的样本中有多少是真正的正样本。[br][br] 对于二分类模型来说,可以直接调用以上公式进行计算;对于多分类模型来说,要用Macro Average 规则来进行计算,即将多分类模型分解为多个二分类模型,分别依据以上公式计算各个精确度,然后取平均值即可。[br][br] 比如上例是个四分类模型,则 acc 类 的 [math]Precision=\frac{106}{130}=81.54\%[/math], good 类 的[br][br][math]Precision=\frac{11}{19}=87.89\%[/math],unacc 类的 [math]Precision=\frac{340}{359}=94.71\%[/math],vgood 类的 [math]Precision=\frac{9}{10}=90\%[/math]。则总的 [math]Precision=\frac{\text{81.54%+ 87.89%+ 94.71%+ 90%}}{4}=88.54\%。[/math][br] [br] (3)recall:称之为召回率。其计算公式为[math]Recall=\frac{TP}{P}[/math],是指正确预测出的正例数占样本中总正例数的比例。实际上,召回率是针对原来的样本而言的,它表示的是样本中的正例有多少被预测正确了。[br][br] (4)F1:预测的调和平均率。其计算公式为 [math]F_1=\frac{2}{\frac{1}{Precision}+\frac{1}{Recall}}=\frac{TP}{TP+\frac{FN+FP}{2}}[/math],从公[br]式中可以看出,F1 是求精准率和召回率的调和平均数的指标。通过对二者求调和平均数,能够更好地反映学习器预测效能的实际平均情况。[br][br] (5)AUC:Area Under Curve 的简称,表示 ROC 曲线下的面积。ROC 是 Receiver Operating Characteristic 的缩写,一般称之为“接受者操作特性曲线”。而 ROC 又与以下两个概念相关:[br][br] [math]\text{TPR=灵敏度(Sensitivity)}=\frac{TP}{TP+FN}[/math][br][br] [math]\text{FPR=1-特异度(Specificity)=}\frac{FP}{FP+TN}[/math][br][br] 在一个二分类模型中,对于不同的阈值,可以得到连续的两值。为了形象化这一变化,在此引入 ROC,ROC 曲线可以用于评价一个分类器。在 Orange 中本例的 ROC 曲线如图 2-4-10 所示。[br][br][center][img]https://s21.ax1x.com/2025/02/12/pEueJN8.png[/img][br] 图 2-4-10 汽车数据案例 ROC 曲线[/center][br] 曲线由两个变量 1-Specificity 和 Sensitivity 绘制,1-Specificity=FPR,即假正类率。Sensitivity 即是真正类率。TPR(True positive rate)反映了正类覆盖程度。这个组合以 1-Specificity 对 Sensitivity,即是以代价(costs)对收益(benefits)。[br][br] 而 AUC 表示 ROC 中曲线下的面积,用于判断模型的优劣。如图 2-4-10 所示,连接对角线的面积刚好是 0.5,对角线的含义也就是随机判断预测结果,正负样本覆盖应该都是 50%。另外,ROC 曲线越陡越好,所以理想值是 1,即正方形。AUC 的值一般介于 0.5 和 1 之间。[br]AUC 评判标准可参考如下:[br][br] ● 0.5~0.7:效果较低。[br][br] ● 0.7~0.85:效果一般。[br][br] ● 0.85~0.95:效果很好。[br][br] ● 0.95~1:效果非常好。[br][br][br]