1+ import sys
2+ import os
3+ path = os .path .abspath (os .path .join ('.' ))
4+ sys .path .append (path )
5+ 6+ import os
7+ import pandas as pd
8+ import seaborn as sns
9+ import numpy as np
10+ import math
11+ from itertools import product , cycle
12+ from sklearn import covariance , preprocessing , tree , svm , neighbors , metrics , linear_model , manifold , linear_model
13+ from sklearn_pandas import DataFrameMapper ,CategoricalImputer
14+ from sklearn import model_selection , ensemble , preprocessing , decomposition , feature_selection
15+ import matplotlib .pyplot as plt
16+ from matplotlib .colors import ListedColormap
17+ from mpl_toolkits .mplot3d import Axes3D
18+ from sklearn .datasets import make_circles , make_moons , make_classification ,make_blobs
19+ import matplotlib .cm as cm
20+ from classification_utils import *
21+ 22+ 23+ def grid_search_plot_one_parameter_curves_clustering (estimator , grid , X , scoring ):
24+ name = str (estimator )
25+ items = sorted (grid .items ())
26+ keys , values = zip (* items )
27+ params = []
28+ scores = []
29+ for v in product (* values ):
30+ params .append (dict (zip (keys , v )))
31+ for param in params :
32+ estimator .set_params (** param )
33+ if (name .startswith ('Gaussian' )):
34+ labels = estimator .fit_predict (X )
35+ else :
36+ estimator .fit (X )
37+ labels = estimator .labels_
38+ 39+ if scoring == 's_score' :
40+ score = metrics .silhouette_score (X , labels , metric = 'euclidean' )
41+ elif scoring == 'ch_score' :
42+ score = metrics .calinski_harabaz_score (X , labels )
43+ else :
44+ print (str (scoring ) + ' metric not supported' )
45+ return
46+ scores .append (score )
47+ 48+ plt .figure ()
49+ plt .plot (params , scores , marker = "D" )
50+ plt .xlabel ('nclusters' )
51+ plt .ylabel (str (scoring ))
52+ 53+ def grid_search_best_model_clustering (estimator , grid , X , scoring ):
54+ name = str (estimator )
55+ items = sorted (grid .items ())
56+ keys , values = zip (* items )
57+ params = []
58+ for v in product (* values ):
59+ params .append (dict (zip (keys , v )))
60+ n = len (params )
61+ best_param = None
62+ best_score = 0.0
63+ for param in params :
64+ estimator .set_params (** param )
65+ if (name .startswith ('Gaussian' )):
66+ labels = estimator .fit_predict (X )
67+ else :
68+ estimator .fit (X )
69+ labels = estimator .labels_
70+ if scoring == 's_score' :
71+ score = metrics .silhouette_score (X , labels , metric = 'euclidean' )
72+ elif scoring == 'ch_score' :
73+ score = metrics .calinski_harabaz_score (X , labels )
74+ else :
75+ print (scoring + ' metric not supported' )
76+ break
77+ if score > best_score :
78+ best_score = score
79+ best_param = param
80+ if best_param is not None :
81+ estimator .set_params (** best_param )
82+ estimator .fit (X )
83+ print ("Best score:" + str (best_score ))
84+ return estimator
85+ else :
86+ return None
87+ 88+ def grid_search_plot_models_2d_clustering (estimator , grid , X , xlim = None , ylim = None ):
89+ plt .style .use ('seaborn' )
90+ items = sorted (grid .items ())
91+ keys , values = zip (* items )
92+ params = []
93+ for v in product (* values ):
94+ params .append (dict (zip (keys , v )))
95+ n = len (params )
96+ fig , axes = plt .subplots (int (math .sqrt (n )), math .ceil (math .sqrt (n )), figsize = (20 , 20 ), dpi = 80 )
97+ axes = np .array (axes )
98+ for ax , param in zip (axes .reshape (- 1 ), params ):
99+ estimator .set_params (** param )
100+ estimator .fit (X )
101+ plot_model_2d_clustering (estimator , X , ax , xlim , ylim , str (param ), False )
102+ plt .tight_layout ()
103+ 104+ def grid_search_plot_models_3d_clustering (estimator , grid , X , xlim = None , ylim = None , zlim = None ):
105+ plt .style .use ('seaborn' )
106+ items = sorted (grid .items ())
107+ keys , values = zip (* items )
108+ params = []
109+ for v in product (* values ):
110+ params .append (dict (zip (keys , v )))
111+ n = len (params )
112+ fig , axes = plt .subplots (int (math .sqrt (n )), math .ceil (math .sqrt (n )), figsize = (20 , 20 ), dpi = 80 , subplot_kw = dict (projection = '3d' ) )
113+ axes = np .array (axes )
114+ for ax , param in zip (axes .reshape (- 1 ), params ):
115+ estimator .set_params (** param )
116+ estimator .fit (X )
117+ plot_model_3d_clustering (estimator , X , ax , xlim , ylim , zlim , str (param ), False )
118+ plt .tight_layout ()
119+ 120+ def plot_model_3d_clustering (estimator , X , ax = None , xlim = None , ylim = None , zlim = None , title = None , new_window = True , rotation = False ):
121+ name = str (estimator )
122+ if (name .startswith ('Gaussian' )):
123+ y = estimator .fit_predict (X )
124+ else :
125+ y = estimator .labels_
126+ 127+ ax = plot_data_3d_classification (X , y , ax , xlim , ylim , zlim , title , new_window , rotation )
128+ if hasattr (estimator , 'cluster_centers_' ):
129+ centers = estimator .cluster_centers_
130+ plot_data_3d (centers , ax , new_window = False , title = title , s = 200 )
131+ 132+ def plot_model_2d_clustering (estimator , X , ax = None , xlim = None , ylim = None , title = None , new_window = True ):
133+ name = str (estimator )
134+ if (name .startswith ('Gaussian' )):
135+ y = estimator .fit_predict (X )
136+ else :
137+ y = estimator .labels_
138+ 139+ ax = plot_data_2d_classification (X , y , ax , xlim , ylim , title , new_window )
140+ if hasattr (estimator , 'cluster_centers_' ):
141+ centers = estimator .cluster_centers_
142+ plot_data_2d (centers , ax , new_window = False , title = title , s = 200 )
143+ 144+ 145+ def generate_synthetic_data_2d_clusters (n_samples , n_centers , cluster_std ) :
146+ return make_blobs (n_samples = n_samples , centers = n_centers ,
147+ cluster_std = cluster_std , random_state = 0 )
148+ 149+ def generate_synthetic_data_3d_clusters (n_samples , n_centers , cluster_std ) :
150+ return make_classification (n_samples = n_samples ,
151+ n_features = 3 ,
152+ n_informative = 3 ,
153+ n_clusters_per_class = 1 ,
154+ n_redundant = 0 ,
155+ n_classes = n_centers ,
156+ random_state = 100 )
0 commit comments