R Tutorial: Hyperparameter tuning in caret

Key Takeaways

carrot makes hyper parameter tuning very easy by default it performs automatic tuning for you with every training run but you can also manually find how you want to tune your models here you see the random forest model with the RF method from before in the output we see that we only have one hyper parameter to tune em try defines the number of variables that are randomly sampled as candidates at each split carrot automatically tried three different my values and includes the performance of each with the output the best model is chosen with the metric accuracy which in this case was 4m try equals six different algorithms have different hyper parameters you might be wondering how you would know which high parameters you can tune with these different methods in carrot if you know the model abbreviation you can use the model lookup function but the easiest way is to use the online documentation of carrot click this link on the slide to go to the page there you will find an overview of the different algorithms you can set as method in the trade function this table includes the name of the model the string you need to put into the train function whether it can be used for classification or regression and what the original are package of the implementation is but most importantly you will find which hyper parameters can be tuned here I will not discuss the mathematics behind hyper parameters instead I will focus on how to perform the high parameter tuning let's change things up a bit and build a support vector machine with polynomial kernel similar to the random forest model from before this time I am using the SVM poly method and I am again calculating the training time when we examined the model object again we see that this time carrot performed a more complex hyper parameter tuning if we have more than one hyper parameter to tune train automatically creates a grid of tuning parameters by default carrot tries all possible combinations of three hyper parameters in our model degree being one two or three scale being 0.01 Oh point O 1 or 0.1 and C being 0.25 0.5 or 1 because the output shows the performance for every possible combination of Hiva parameters the output is too long to fit on this night and I am only showing the best model with degree of 1 scale of 0.1 and C equal to 1 we can also set the option tune lengths to specify the number of different values to try for each hyper parameter for example 5 now carat tries all possible combinations of 5 5 parameters degree B 1 2 3 4 or 5 scale being 1 to the power of minus 3 1 to the power of minus 2 1 to the power of minus 1 1 and 10 and C being 0.25 0.5 1 2 or 4 the best model now has degree scale and C of 1 of course you could also manually try out different hyper parameters this we can do with the option to in grid 2 which we can feed a grid of hyper parameters this grade is defined with the expanded grade function if we use that function we need to define all height parameters let's see what happens if we set the degree to 4 and keep scale and see it 1 and 3 train the model this time we only train with 1 combination of 1/2 parameters so our output gives the performance for these hyper parameters only now's your turn to apply simple height parameter