-
-
Notifications
You must be signed in to change notification settings - Fork 165
/
utils.py
172 lines (150 loc) · 6.99 KB
/
utils.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
import numpy as np
import pandas as pd
from . import checks
from .. import _global_checks
def calculate_cat_assoc_matrix(data, categorical_variables, n):
from scipy.stats import chi2_contingency
cat_assoc_matrix = pd.DataFrame(
index=categorical_variables, columns=categorical_variables, dtype=float
)
for i, variable_name_a in enumerate(categorical_variables):
for j, variable_name_b in enumerate(categorical_variables):
if i > j:
continue # matrix is symmetric
elif i == j:
cramers_V = 1
else:
# calculate Cramér’s V with bias correction
cont_tab = pd.crosstab(data[variable_name_a], data[variable_name_b])
r = cont_tab.shape[0]
c = cont_tab.shape[1]
chi2_test = chi2_contingency(cont_tab)
phi = max(0, (chi2_test[0] / n) - (((r - 1) * (c - 1)) / (n - 1)))
r = r - ((r - 1) ** 2 / (n - 1))
c = c - ((c - 1) ** 2 / (n - 1))
cramers_V = np.sqrt(phi / (min(c, r) - 1))
cramers_V = checks.check_assoc_value(cramers_V)
cat_assoc_matrix.loc[variable_name_a, variable_name_b] = cramers_V
cat_assoc_matrix.loc[variable_name_b, variable_name_a] = cramers_V
return cat_assoc_matrix
def calculate_cat_num_assoc_matrix(data, categorical_variables, numerical_variables, n):
from scipy.stats import kruskal
cat_num_assoc_matrix = pd.DataFrame(
index=categorical_variables, columns=numerical_variables, dtype=float
)
for cat_variable_name in categorical_variables:
for num_variable_name in numerical_variables:
# calculate eta-squared
samples = [
gr_data[num_variable_name].values
for gr_name, gr_data in data.groupby(cat_variable_name, as_index=False)
]
kruskal_wallis_test = kruskal(*samples, nan_policy="omit")
n_samples = len(samples)
eta_squared = (kruskal_wallis_test.statistic - n_samples + 1) / (n - n_samples)
eta_squared = checks.check_assoc_value(eta_squared)
cat_num_assoc_matrix.loc[cat_variable_name, num_variable_name] = eta_squared
return cat_num_assoc_matrix
def calculate_assoc_matrix(data, corr_method):
corr_matrix = abs(data.corr(corr_method)) # get submatrix for only numerical variables
numerical_variables = corr_matrix.columns
categorical_variables = list(set(data.columns) - set(numerical_variables))
n = len(data)
cat_assoc_matrix = calculate_cat_assoc_matrix(data, categorical_variables, n) # get submatrix for only categorical variables
cat_num_assoc_matrix = calculate_cat_num_assoc_matrix( # get submatrix for mixed variables
data, categorical_variables, numerical_variables, n
)
tmp_matrix = pd.concat([corr_matrix, cat_num_assoc_matrix.T], axis=1)
tmp_matrix_2 = pd.concat([cat_num_assoc_matrix, cat_assoc_matrix], axis=1)
assoc_matrix = pd.concat([tmp_matrix, tmp_matrix_2])
return assoc_matrix
def calculate_pps_matrix(data, agg_method):
_global_checks.global_check_import('ppscore', 'depend_matrix with PPS calculation')
import ppscore as pps
pps_result = pps.matrix(data, sample=None)
pps_matrix = pps_result[["x", "y", "ppscore"]].pivot(
columns="x", index="y", values="ppscore"
)
pps_matrix.rename_axis(None, axis=1, inplace=True)
pps_matrix.rename_axis(None, axis=0, inplace=True)
# aggregate values to make symmetric matrix
if agg_method == "max":
pps_matrix = np.maximum(pps_matrix, pps_matrix.transpose())
if agg_method == "min":
pps_matrix = np.minimum(pps_matrix, pps_matrix.transpose())
if agg_method == "mean":
pps_matrix = (pps_matrix + pps_matrix.transpose()) / 2
return pps_matrix
def calculate_depend_matrix(
data, depend_method, corr_method, agg_method
):
depend_matrix = pd.DataFrame()
if depend_method == "assoc":
depend_matrix = calculate_assoc_matrix(data, corr_method)
if depend_method == "pps":
depend_matrix = calculate_pps_matrix(data, agg_method)
if callable(depend_method):
try:
depend_matrix = depend_method(data)
except:
raise ValueError(
"You have passed wrong callable in depend_method argument. "
"'depend_method' is the callable to use for calculating dependency matrix."
)
return depend_matrix
def calculate_linkage_matrix(depend_matrix, clust_method):
from scipy.cluster.hierarchy import linkage
from scipy.spatial.distance import squareform
dissimilarity = 1 - abs(depend_matrix)
## https://www.kaggle.com/sgalella/correlation-heatmaps-with-hierarchical-clustering?scriptVersionId=24045077&cellId=10
linkage_matrix = linkage(squareform(dissimilarity), clust_method)
return linkage_matrix
def get_dendrogram_aspects_ordered(hierarchical_clustering_dendrogram, depend_matrix):
# names of variables on axis
tick_dict = dict(
zip(
hierarchical_clustering_dendrogram.layout.yaxis.tickvals,
[[var] for var in hierarchical_clustering_dendrogram.layout.yaxis.ticktext]
)
)
d = {k:v for v,k in enumerate(depend_matrix.columns)} # original order of columns
# get names of variables for dendrogram traces
_aspects_dendrogram_order = []
for scatter in hierarchical_clustering_dendrogram.data:
vars_list = tick_dict[scatter.y[1]] + tick_dict[scatter.y[2]]
vars_list.sort(key=d.get)
tick_dict[np.mean(scatter.y[1:3])] = vars_list
_aspects_dendrogram_order.append(
vars_list
)
# get min dependency between grouped variables
_vars_min_depend, _min_depend = get_min_depend_from_matrix(depend_matrix, _aspects_dendrogram_order)
return pd.DataFrame({
"variable_names": _aspects_dendrogram_order,
"min_depend": _min_depend,
"vars_min_depend": _vars_min_depend
})
def get_min_depend_from_matrix(depend_matrix, variables_list):
_vars_min_depend = []
_min_depend = []
for vars in variables_list:
if isinstance(vars, (list, tuple, np.ndarray)):
depend_submatrix = abs(depend_matrix).loc[list(vars), list(vars)]
_min = np.array(depend_submatrix).min()
_min_idx = np.where(depend_submatrix == _min)
var_a_idx, var_b_idx = _min_idx[0][0], _min_idx[1][0]
_vars_min_depend.append(np.array(depend_submatrix.columns[[var_a_idx, var_b_idx]]))
_min_depend.append(_min)
else:
_vars_min_depend.append(None)
_min_depend.append(None)
return _vars_min_depend, _min_depend
def calculate_min_depend(
variables_list,
data,
depend_method="assoc",
corr_method="spearman",
agg_method="max",
):
depend_matrix = calculate_depend_matrix(data, depend_method, corr_method, agg_method)
return get_min_depend_from_matrix(depend_matrix, variables_list)