cluster
¶Fuzzy clustering subpackage, containing fuzzy cmeans clustering algorithm. This can be either supervised or unsupervised, depending if U_init kwarg is used (if guesses are provided, it is supervised).
skfuzzy.cluster.cmeans (data, c, m, error, ...) 
Fuzzy cmeans clustering algorithm [1]. 
skfuzzy.cluster.cmeans_predict (test_data, ...) 
Prediction of new data in given a trained fuzzy cmeans framework [1]. 
skfuzzy.cluster.
cmeans
(data, c, m, error, maxiter, init=None, seed=None)[source]¶Fuzzy cmeans clustering algorithm [1].
Parameters:  data : 2d array, size (S, N)
c : int
m : float
error : float
maxiter : int
init : 2d array, size (S, N)
seed : int


Returns:  cntr : 2d array, size (S, c)
u : 2d array, (S, N)
u0 : 2d array, (S, N)
d : 2d array, (S, N)
jm : 1d array, length P
p : int
fpc : float

Notes
The algorithm implemented is from Ross et al. [R24].
Fuzzy CMeans has a known problem with high dimensionality datasets, where the majority of cluster centers are pulled into the overall center of gravity. If you are clustering data with very high dimensionality and encounter this issue, another clustering method may be required. For more information and the theory behind this, see Winkler et al. [R25].
References
[R24]  (1, 2) Ross, Timothy J. Fuzzy Logic With Engineering Applications, 3rd ed. Wiley. 2010. ISBN 9780470743768 pp 352353, eq 10.28  10.35. 
[R25]  (1, 2) Winkler, R., Klawonn, F., & Kruse, R. Fuzzy cmeans in high dimensional spaces. 2012. Contemporary Theory and Pragmatic Approaches in Fuzzy Computing Utilization, 1. 
skfuzzy.cluster.
cmeans_predict
(test_data, cntr_trained, m, error, maxiter, init=None, seed=None)[source]¶Prediction of new data in given a trained fuzzy cmeans framework [1].
Parameters:  test_data : 2d array, size (S, N)
cntr_trained : 2d array, size (S, c)
m : float
error : float
maxiter : int
init : 2d array, size (S, N)
seed : int


Returns:  u : 2d array, (S, N)
u0 : 2d array, (S, N)
d : 2d array, (S, N)
jm : 1d array, length P
p : int
fpc : float

Notes
Ross et al. [R26] did not include a prediction algorithm to go along with fuzzy cmeans. This prediction algorithm works by repeating the clustering with fixed centers, then efficiently finds the fuzzy membership at all points.
References
[R26]  (1, 2) Ross, Timothy J. Fuzzy Logic With Engineering Applications, 3rd ed. Wiley. 2010. ISBN 9780470743768 pp 352353, eq 10.28  10.35. 