mirror of
https://github.com/meteoinfo/MeteoInfo.git
synced 2025-12-08 20:36:05 +00:00
134 lines
3.6 KiB
Python
134 lines
3.6 KiB
Python
from org.meteoinfo.math.stats import StatsUtil, Jenks
|
|
|
|
from ..core import numeric as np
|
|
|
|
__all__ = ["jenks_breaks", "JenksNaturalBreaks"]
|
|
|
|
|
|
class JenksNaturalBreaks(object):
|
|
"""
|
|
A class that can be used to classify a sequence of numbers into groups (clusters) using Fisher-Jenks natural breaks.
|
|
"""
|
|
|
|
def __init__(self, n_classes=6):
|
|
"""
|
|
Parameters
|
|
----------
|
|
n_classes : int
|
|
The number of classes to be generated by the classifier.
|
|
"""
|
|
self.n_classes = n_classes
|
|
|
|
def __repr__(self):
|
|
return "JenksNaturalBreaks(n_classes={})".format(self.n_classes)
|
|
|
|
def __str__(self):
|
|
return self.__repr__()
|
|
|
|
def fit(self, x):
|
|
"""
|
|
Parameters
|
|
----------
|
|
x : array-like
|
|
The sequence of numbers (integer/float) to be classified.
|
|
"""
|
|
x = np.asarray(x)
|
|
|
|
self._jenks = Jenks()
|
|
self._jenks.addValues(x.asarray())
|
|
self._breaks = self._jenks.computeBreaks(self.n_classes)
|
|
self.breaks_ = np.array(self._breaks.getClassValues())
|
|
self.inner_breaks_ = self.breaks_[1:-1] # because inner_breaks is more
|
|
self.labels_ = self.predict(x)
|
|
self.groups_ = self.group(x)
|
|
|
|
def predict(self, x):
|
|
"""
|
|
Predicts the class of each element in x.
|
|
|
|
Parameters
|
|
----------
|
|
x : scalar or array-like
|
|
|
|
Returns
|
|
-------
|
|
array
|
|
"""
|
|
x = np.asarray(x)
|
|
r = self._breaks.classOf(x.asarray())
|
|
return np.array(r)
|
|
|
|
def group(self, x):
|
|
"""
|
|
Groups the elements in x into groups according to the classifier.
|
|
|
|
Parameters
|
|
----------
|
|
x : array-like
|
|
The sequence of numbers (integer/float) to be classified.
|
|
|
|
Returns
|
|
-------
|
|
list of numpy.array
|
|
The list of groups that contains the values of x.
|
|
"""
|
|
arr = np.array(x)
|
|
groups_ = [arr[arr <= self.inner_breaks_[0]]]
|
|
for idx in range(len(self.inner_breaks_))[:-1]:
|
|
groups_.append(arr[(arr > self.inner_breaks_[idx]) & (arr <= self.inner_breaks_[idx + 1])])
|
|
groups_.append(arr[arr > self.inner_breaks_[-1]])
|
|
return groups_
|
|
|
|
def goodness_of_variance_fit(self, x):
|
|
"""
|
|
Parameters
|
|
----------
|
|
x : array-like
|
|
|
|
Returns
|
|
-------
|
|
float
|
|
The goodness of variance fit.
|
|
"""
|
|
gvf = self._breaks.gvf()
|
|
return gvf
|
|
|
|
def get_label_(self, val):
|
|
"""
|
|
Compute the group label of the given value.
|
|
|
|
Parameters
|
|
----------
|
|
val : float
|
|
The value to be classified.
|
|
|
|
Returns
|
|
-------
|
|
int : The label of the value.
|
|
"""
|
|
return self._breaks.classOf(val)
|
|
|
|
|
|
def jenks_breaks(a, n_classes, gvf=False):
|
|
"""
|
|
Compute natural breaks (Fisher-Jenks algorithm) on a sequence of `values`,
|
|
given `n_classes`, the number of desired class.
|
|
|
|
:param a: (*array*) Input data array.
|
|
:param n_classes: (*int*) The number of desired class.
|
|
:param gvf: (*float*) Whether return gvf (Goodness of Variance Fit) value. Default is `False`.
|
|
|
|
:return: (*array*) The computed break values, including minimum and maximum, in order
|
|
to have all the bounds for building `n_classes` classes,
|
|
so the returned list has a length of `n_classes` + 1.
|
|
"""
|
|
if isinstance(a, (list, tuple)):
|
|
a = np.array(a)
|
|
|
|
if gvf:
|
|
r = StatsUtil.jenksBreaksGvf(a._array, n_classes)
|
|
return np.array(r[0]), r[1]
|
|
else:
|
|
r = StatsUtil.jenksBreaks(a._array, n_classes)
|
|
return np.array(r)
|