基于t-SNE的随机森林可视化

VISUALIZING RANDOM FOREST BASED ON T-SNE

随机森林是一种优秀的组合分类器,但缺少较好的解释性。为了使随机森林模型更具理解性和解释性,本文提出一种基于t-SNE的可视化随机森林相似性矩阵的方法:首先运用随机森林学习出样本间的相似性度量矩阵,然后采用t-SNE方法降维,最后可视化。实验证明,该方法比MDS更有效。

Random forest is an excellent ensemble classifier, but it is the lack of interpretability,whmake the random forest model more understandable and interpretable, we put forward a method based on t-SNE. Firstly, we use of ran-dom forest to learn a similarity measiare between the sample, then dimension reduction and visualization by the method of t-SNE. The experiment result shows that the proposed method is more effective than MDS.