What does get_fscore() of an xgboost ML model do?

Question

Does anybody how the numbers are calculated? In the documentation it says that this function "Get feature importance of each feature", but there is no explanation on how to interpret the results.

Answer 1

This is a metric that simply sums up how many times each feature is split on. It is analogous to the Frequency metric in the R version.https://cran.r-project.org/web/packages/xgboost/xgboost.pdf

It is about as basic a feature importance metric as you can get.

i.e. How many times was this variable split on?

The code for this method shows it is simply adding of the presence of a given feature in all the trees.

[here..https://github.com/dmlc/xgboost/blob/master/python-package/xgboost/core.py#L953][1]

def get_fscore(self, fmap=''):
    """Get feature importance of each feature.
    Parameters
    ----------
    fmap: str (optional)
       The name of feature map file
    """
    trees = self.get_dump(fmap)  ## dump all the trees to text
    fmap = {}                    
    for tree in trees:              ## loop through the trees
        for line in tree.split('\n'):     # text processing
            arr = line.split('[')
            if len(arr) == 1:             # text processing 
                continue
            fid = arr[1].split(']')[0]    # text processing
            fid = fid.split('<')[0]       # split on the greater/less(find variable name)

            if fid not in fmap:  # if the feature id hasn't been seen yet
                fmap[fid] = 1    # add it
            else:
                fmap[fid] += 1   # else increment it
    return fmap                  # return the fmap, which has the counts of each time a  variable was split on