Unverified50% confidenceFactExact time
决策树及其集成版本（随机森林、XGBoost）对特征缩放不敏感，不需要像SVM或神经网络那样进行标准化预处理
1
Sources
50%
Confidence
Long-term
Relevance
8/24/2026
First Seen
Sources
Related Entities
Related Claims
Unverified该SWD特征变换实验在决策树上表现良好，但在逻辑回归、SVM、KNN等算法上效果糟糕66% similarUnverified决策树使用沿坐标轴的轴对齐分割，对特征的单调变换和局部拉伸不敏感，因此对保持类别排序的变换具有天然不变性63% similarUnverified当数据包含大量高基数类别特征时，基于距离的算法（如K-means、KNN）表现不佳，基于树的模型（如随机森林、梯度提升树）通常更稳妥61% similarUnverifiedUnsupervised learning does not rely on labels and attempts to discover inherent structure in data, such as clustering and dimensionality reduction.57% similarUnverifiedscikit-learn 0.22 将决策树的默认分裂标准从 'mse' 更名为 'squared_error'56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/793942API
curl https://kongchang.com/api/v1/knowledge/claims/793942MCP
get_claim(id=793942)