Python implementation of the Gap Statistic

Forked from original repository

---

Purpose

Dynamically identify the suggested number of clusters in a data-set using the gap statistic.

---

Improvements

• Correct dispersion formula (mean of log instead of log of mean)
• Compute gap statistic's standard deviation
• Add Scikit-learn KMeans and SphericalKMeans
• Scipy kmeans2 looks very unstable, that's why it's not the default algorithm anymore

---

Full example available in a notebook HERE

---

Install:

Bleeding edge:

pip install git+git://github.com/druogury/clustering-gap-statistic.git

PyPi:

pip install --upgrade gap-stat

---

Uninstall:

pip uninstall gap-stat