2

Is there the library in R, which work in the next way? 1. load the data! 2 it iterates model using different methods (neural network, random forest,(g)lm, etc) and different setting until it gets the best results which is possible on this data? Something like is implemented in SPSS Modeler , but works very bad! Maybe this feature there is in R or another language?

schoolboy
  • 21
  • 1
  • 1
    It [generally works bad](https://stats.stackexchange.com/questions/20836/algorithms-for-automatic-model-selection/20856#20856) – Tim Aug 20 '17 at 15:30
  • 1
    Possible duplicate of [Algorithms for automatic model selection](https://stats.stackexchange.com/questions/20836/algorithms-for-automatic-model-selection) – Tim Aug 20 '17 at 15:30
  • o' right, suppose , i build any models (logistic regression and neural network) and they work bad . i explain! I have dataset - one dep .var. (o or 1) and 4 scale vars. And the result that, all obs. were classed as 0. What i should to do? I can't say to my colleague something like "on this data we can't build the models, sorry, where my money". I must get the results!!! – schoolboy Aug 20 '17 at 15:44
  • well one feature that goes in this direction is the auto.arima() function from the forecast package – Ferdi Aug 20 '17 at 15:56
  • yes, i know it) Is thete auto.nnet?:) – schoolboy Aug 20 '17 at 16:39
  • @schoolboy there is a difference between "nothing can be done" and deciding to use a method that makes your results prone to overfitting. What the linked thread shows is that such "automatic model selection" procedures make your results prone to overfitting. Moreover, the main core of your question, i.e. asking for code in R, is off-topic in here. – Tim Aug 20 '17 at 17:05
  • Tim, not correct! It's not obligatory R, I will also be satisfied with theoretical reflections, cause there many methods and it's settings, i just don't want do it manually. I am sure that the way of automatic choosing must be exist! – schoolboy Aug 20 '17 at 19:15

0 Answers0