<p>Bagging is the ensembl modeling. Bagging combines the several models and take the&nbsp;weighted or unweighted average, voting estimation of&nbsp;all models.&nbsp; The more different models are, the more powerful the Bagging is. Therefore, Bagging is normally based on Decision Tree, which is unstable modeling method.</p><p>In your output, Bagging generates 10 trees(assuming you use tree), and n, min, max , mean show the statisitics of 10 trees. , i.e the minimum misclassification rate for validation dataset is 0.1746. </p><p>Testing dataset is optional for model buidling because validation dataset can help you tune and select the optimal model. However, if you want to verify your model then you can use testing dataset to get the unbiased prediction to check it. Particularly, when your dataset is small, it is not wise to partition part of your dataset as testing dataset.</p><p>I have one question, are you using EM 4.0 to get bagging? or EM 5.3. I am sure in Em5.2, there is no bagging, boosting and randomforest.</p>

<p>Hi, thanks for your answer. I'm using EM4.3. by the way, where can I finding the&nbsp;final classification error&nbsp;from bagging? It seems that it should take majority voting as the final prediction result. Does it show&nbsp;as the classification table in my output ? Also, do you know how to do bootstrap for decision tree using EM ? Is bootstrap and cross validation used for model selection also? </p>

路過 學習一下

学习了