Bagging model

User13User13 Member Posts: 155 Maven
edited November 2018 in Help
Hi

How can in RapidMiner software,convert the outputs of the hybrid model ( eg Bagging) to a final model?

Thanks

Answers

  • User13User13 Member Posts: 155 Maven
    Hi Eli,

    the model output of the Bagging operator provides a model which you can use like any other model in RapidMiner Studio, there is no need to combine single models to one object. This ensemble model contains a fixed number of models. When you apply the ensemble model all inner models will be applied and a prediction is computed i.e. via majority voting. If you want to combine different models originating from different modeling operators you can use the operator "Vote" to train an ensemble model holding different types of models.

    Cheers,
    Helge
  • User13User13 Member Posts: 155 Maven
    How does your process look like? Have you tried to use the "Bagging" operator? This will automatically do the process for you and combine your models to one final model object.
  • User13User13 Member Posts: 155 Maven
    I want output analysis. if number of iteration = 3 and learning process = decision tree ,then Will be seen 3 tree in output. which of the 3 output models must be analyze?
  • User13User13 Member Posts: 155 Maven
    Here, there is no possibility of the attached file. You may put your email to send my files?
  • User13User13 Member Posts: 155 Maven
    You may paste your XML code in here or send an email to me (please check your inbox).
  • User13User13 Member Posts: 155 Maven
    You have several separate models because you are using the bagging feature. This feature trains several models of the inner learner (here a decision tree) and combines them to an ensemble model. You need to use this feature in case your learner tends to become unstable when trained with different samples of your data. The model shown in your last picture actually consists of three decision tree models. If you want to train a single decision tree model please remove the bagging operator and replace it directly with the decision tree learner. You may also want to compare the results for your process with and without bagging to check whether you really need the bagging approach.
Sign In or Register to comment.