Diffusion-weighted Magnetic Resonance Imaging (dMRI) can unveil the microstructure of the brain white matter. The analysis of the anisotropy observed in the dMRI and highlighted by tractography methods can help to understand the pattern of connections between brain regions and characterize neurological diseases. Because of the amount of information produced by such analyses and the errors carried by the reconstruction step, it is necessary to simplify this output. Clustering algorithms can be used to group samples that are similar according to a given metric. In this work, we address important questions related to the use of clustering for brain fibers obtained by dMRI, namely: i) what is the most adequate metric to quantify the similarity between brain fiber tracts? ii) How to select the best clustering algorithm and parametrization among standard possibilities? While trying to solve these questions, we perform a new contribution: we show how to combine the well-known K-means clustering algorithm with various metrics while keeping an efficient procedure. We analyze the performance and usability of the ensuing algorithms on a dataset of ten subjects. We show that the association of K-means with Point Density Model, a recently proposed metric to analyze geometric structures, outperforms other state-of-the-art solutions.