三、KNN 的交叉验证

[br] 由于 KNN 算法本身具有动态调整的特点,需要不断调整测试集和训练集的比例才有较为良好的模型表现,所以交叉验证在 KNN 算法的应用过程中显得尤为重要。一般来讲,按比例的数据采样或多折交叉验证需要多次应用,相互印证。[br][br] 在机器学习里,通常不能将全部数据用于训练模型,否则将没有数据集对该模型进行验证,从而难以评估模型的预测效果。为了解决这一问题,通常将数据集进行不同方法的分割,人为将整个数据集分为训练集(training set)和测试集(test set)两部分。我们将这种数据集分割组合进行算法模型评估的方法叫作交叉验证。示意图如下:[br][center][img]https://s21.ax1x.com/2025/02/13/pEu0Q76.jpg[/img][/center] 在实际应用中,有如下几种常用方法:
[br][b] (一)The Validation Set Approach(验证集方法)[br][/b][br] 该方法是把整个数据集分成两部分,一部分用于训练,另一部分用于验证,分割时以比例进行调节。[br][br] 同样以汽车分类为例,以不同的比例分割数据集,其错误率是不同的,如图 2-4-4 所示。[br][center][img]https://s21.ax1x.com/2025/02/12/pEuVIMR.jpg[/img][/center][center]图 2-4-4 汽车分类案例不同比例分割的错误率[/center][br] 可以看到,在不同的划分方法下,错误率的变动是很大的。所以如果训练集和测试集的划分方法不够好,很有可能无法选择到最好的模型与参数。
[br][b] (二)Cross-Validation(交叉验证法)[br][/b][br] 1. LOOCV [br][br] LOOCV 方法即 leave-one-out cross-validation,或称为留一法。像 test set approach 一样,LOOCV 方法也包含将数据集分为训练集和测试集这一步骤。但是不同的是,现在只用一个数据作为测试集,其他数据都作为训练集,并将此步骤重复 n 次(n 为数据集的数据数量)。留一法的示意图如下所示:[br][center][img]https://s21.ax1x.com/2025/02/13/pEu0XHx.jpg[/img][/center][br] 假设现在有 n 个数据组成的数据集,那么 LOOCV 方法就是每次取出一个数据作为测试集的唯一元素,而其他 n -1个数据都作为训练集用于训练模型和调参。结果就是最终训练了 n 个模型,每次都能得到一个 MSE(均方误差)。而计算最终 test MSE 则就是将这 n 个 MSE取平均。[br] [math]CV_{\left(n\right)}=\frac{1}{n}\sum^n_{i=1}MSE_i=\frac{1}{n}\sum^n_{i-1}\left(y^i_{true}-y\text{^}\text{ }_{test}^i\right)^2[/math](适用于回归问题)[br] [br] [math]CV\left(n\right)=\frac{1}{n}\sum^n_{i=1}Err_i[/math](适用于分类问题, Err 是指分类错误的数量)[br][br][br] 2. K-fold Cross Validation(K 折交叉验证)[br][br] K折交叉验证的示意图如下所示:[br][center][img]https://s21.ax1x.com/2025/02/13/pEu0i7V.jpg[/img][/center] K 折交叉验证与 LOOCV 的不同之处在于,每次测试集将不再只包含一个数据,而是多个,具体数目将根据 K 的选取决定。比如,如果 K=5,那么利用五折交叉验证的步骤就是:[br] (1)将所有数据集分成 5 份。[br] (2)不重复地每次取其中一份做测试集,用其他 4 份做训练集训练模型,之后计算该模型在测试集上的 MSEi 。[br] (3)将 5 次的 MSEi 取平均得到最后的 MSEi 。[br] 以 Orange 平台为例,有两个模块是被经常用于交叉验证的,说明如下:[br] [br] 1)数据采样器模块[br] 数据采样器模块的名称为 Data Sampler,图标形式是[img]https://s21.ax1x.com/2025/02/12/pEuZ6ld.png[/img] ,在这个模块中,采样类型被分为 4 种,其窗口如图 2-4-5 所示。[br][br] [img]data:image/png;base64,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[/img] [br] 图 2-4-5 数据采样器模块窗口[br][br] ▶Fixed proportion of data:这是一种按固定比例将数据进行分割的方法,其数量可以按照比例条进行调节,一般按 6∶4 或 7∶3 的比例进行分割。[br] ▶ Fixed sample size:这是按照固定数量将数据集进行分割的方法,勾选 Sample with replacement 表示有放回的采样方式。[br] ▶ Cross validation:交叉验证,是常用的一种采样方式,是将数据集平均分为几折,然后取其中的一折作为测试集的验证方法。调节参数 Number of folds 表示分为几折,即几份;Selected fold 表示取那一份为测试集。[br] ▶ Bootstrap:是通过样本来估计总体的采样方法。[br] 可选项 Replicable(deterministic)sampling 是可复制(确定性)采样,Stratify sample 是根据数据属性不同而采取的分层采样。[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] 。在这一模块中,也有采用的参数调节选项,如图 2-4-6 所示。[br] ▶ Cross validation:交叉验证。[br] ▶ Random sampling:随机采样。其中,Repeat train/test 表示重复训练及测试的次数,Training set size 表示随机抽取的训练测试集的比例。[br][br][center][/center][img]data:image/png;base64,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[/img] [br] [center] 图 2-4-6 测试模块窗口[/center][br][br] ▶ Leave one out:表示留一法。[br] ▶ Test on train data:用训练数据测试。[br] ▶Test on test data:用测试数据测试,如数据已经被采样。

Information: 三、KNN 的交叉验证