A recent discretization method for long-tailed distribution (skewed to the left, with very low values and some very strong values). It creates a threshold to the average of the variable, eliminates all individuals below the threshold and creates a new threshold for the data located above the first threshold. The process is repeated until there are no more than a point above the final threshold.

See: https://sites.google.com/site/thepowerofcartography/head-tail-breaks
[French: Head/Tail Breaks Classes]