<blockquote>
<p>Am I right to say that also Standardization gets affected negatively by the extreme values as well? </p>
</blockquote>
<p>事实上,你是;scikit学习<a href="http://scikit-learn.org/0.18/auto_examples/preprocessing/plot_robust_scaling.html" rel="noreferrer">docs</a>他们自己清楚地警告这种情况:</p>
<blockquote>
<p>However, when data contains outliers, <a href="http://scikit-learn.org/0.18/modules/generated/sklearn.preprocessing.StandardScaler.html#sklearn.preprocessing.StandardScaler" rel="noreferrer"><code>StandardScaler</code></a> can often be mislead. In such cases, it is better to use a scaler that is robust against outliers.</p>
</blockquote>
<p>或多或少,对于<code>MinMaxScaler</code>也是如此。</p>
<blockquote>
<p>I really can't see how the <strong><em>Robust Scaler</em></strong> improved the data because I still have <strong><em>extreme values</em></strong> in the resulted data set? Any simple -complete interpretation?</p>
</blockquote>
<p><strong>健壮并不意味着<em>免疫</em>,或<em>不受攻击</em></strong>,缩放的目的是<em>不</em>以“删除”异常值和极值-这是一个单独的任务,有自己的方法;这在<a href="http://scikit-learn.org/stable/auto_examples/preprocessing/plot_all_scaling.html#robustscaler" rel="noreferrer">relevant scikit-learn docs</a>中再次明确提到:</p>
<blockquote>
<p><strong>RobustScaler</strong></p>
<p>[...] Note that the outliers themselves are still present in the transformed data. If a separate outlier clipping is desirable, a non-linear transformation is required (see below).</p>
</blockquote>
<p>其中“see below”指的是<a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.QuantileTransformer.html#sklearn.preprocessing.QuantileTransformer" rel="noreferrer">^{<cd2>}</a>和<a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.quantile_transform.html" rel="noreferrer">^{<cd3>}</a>。</p>