I am doing the prediction modeling, i want to use the cross-validation method and resample method to compare the model prediction accuracy. The model was built by SVM, Random Forrest, and Logistic regression. I used R can not build the resample dataset(bootstrap method ), How can I do it? Thanks for your help. I tried the boot function but can not build the dataset.
Hello Pei,
A simple way to do this is:
Let
nc = number of bootstrap cycles you wish to run
N = number of cases in your sample data set
x = vector (or other structure) containing the sample data set
y = temporary vector (or other structure, same as the sample data set)
z = results vector (or other suitable storage for recording method outcomes)
Loop through the following steps for i = 1 to nc times:
1. Randomly sample N cases from x, with replacement.
2. Assign this sample to row i of the temporary storage, y.
3. Apply a target method to this bootstrap sample in y.
4. Add the result(s) of this cycle as row i of z (which will ultimately have length nc).
The outcome vector should have a set of nc results, summarizing the outcomes of the target method applied to (nc) bootstrap samples. Summarize these as appropriate.
All this is simple to do in R.
For example, here's a simple code for steps 1 and 2, above:
#random sampling with replacement of N cases from vector, x, saved to temp vector, y
#
y = sample(x[1:N], size = N, replace=TRUE)
#
Good luck with your work.
- Badges
- Science topic