|
1
|
|
|
""" |
|
2
|
|
|
Functions for descriptive analytics. |
|
3
|
|
|
|
|
4
|
|
|
:author: Andreas Kanz |
|
5
|
|
|
|
|
6
|
|
|
""" |
|
7
|
|
|
|
|
8
|
|
|
# Imports |
|
9
|
|
|
from matplotlib.colors import LinearSegmentedColormap, to_rgb |
|
10
|
|
|
import matplotlib.pyplot as plt |
|
11
|
|
|
import matplotlib.ticker as ticker |
|
12
|
|
|
import numpy as np |
|
13
|
|
|
import pandas as pd |
|
14
|
|
|
import scipy |
|
15
|
|
|
import seaborn as sns |
|
16
|
|
|
|
|
17
|
|
|
from typing import Any, Dict, Optional, Tuple, Union |
|
18
|
|
|
from klib.utils import ( |
|
19
|
|
|
_corr_selector, |
|
20
|
|
|
_missing_vals, |
|
21
|
|
|
_validate_input_bool, |
|
22
|
|
|
_validate_input_int, |
|
23
|
|
|
_validate_input_range, |
|
24
|
|
|
_validate_input_smaller, |
|
25
|
|
|
_validate_input_sum_larger, |
|
26
|
|
|
) |
|
27
|
|
|
|
|
28
|
|
|
|
|
29
|
|
|
__all__ = ["cat_plot", "corr_mat", "corr_plot", "dist_plot", "missingval_plot"] |
|
30
|
|
|
|
|
31
|
|
|
|
|
32
|
|
|
# Functions |
|
33
|
|
|
|
|
34
|
|
|
# Categorical Plot |
|
35
|
|
|
def cat_plot( |
|
36
|
|
|
data: pd.DataFrame, |
|
37
|
|
|
figsize: Tuple = (18, 18), |
|
38
|
|
|
top: int = 3, |
|
39
|
|
|
bottom: int = 3, |
|
40
|
|
|
bar_color_top: str = "#5ab4ac", |
|
41
|
|
|
bar_color_bottom: str = "#d8b365", |
|
42
|
|
|
# cmap: str = "BrBG", |
|
43
|
|
|
): |
|
44
|
|
|
""" Two-dimensional visualization of the number and frequency of categorical features. |
|
45
|
|
|
|
|
46
|
|
|
Parameters |
|
47
|
|
|
---------- |
|
48
|
|
|
data : pd.DataFrame |
|
49
|
|
|
2D dataset that can be coerced into Pandas DataFrame. If a Pandas DataFrame is provided, the index/column \ |
|
50
|
|
|
information is used to label the plots |
|
51
|
|
|
figsize : Tuple, optional |
|
52
|
|
|
Use to control the figure size, by default (18, 18) |
|
53
|
|
|
top : int, optional |
|
54
|
|
|
Show the "top" most frequent values in a column, by default 3 |
|
55
|
|
|
bottom : int, optional |
|
56
|
|
|
Show the "bottom" most frequent values in a column, by default 3 |
|
57
|
|
|
bar_color_top : str, optional |
|
58
|
|
|
Use to control the color of the bars indicating the most common values, by default "#5ab4ac" |
|
59
|
|
|
bar_color_bottom : str, optional |
|
60
|
|
|
Use to control the color of the bars indicating the least common values, by default "#d8b365" |
|
61
|
|
|
cmap : str, optional |
|
62
|
|
|
The mapping from data values to color space, by default "BrBG" |
|
63
|
|
|
|
|
64
|
|
|
Returns |
|
65
|
|
|
------- |
|
66
|
|
|
Gridspec |
|
67
|
|
|
gs: Figure with array of Axes objects |
|
68
|
|
|
""" |
|
69
|
|
|
|
|
70
|
|
|
# Validate Inputs |
|
71
|
|
|
_validate_input_int(top, "top") |
|
72
|
|
|
_validate_input_int(bottom, "bottom") |
|
73
|
|
|
_validate_input_range(top, "top", 0, data.shape[1]) |
|
74
|
|
|
_validate_input_range(bottom, "bottom", 0, data.shape[1]) |
|
75
|
|
|
_validate_input_sum_larger(1, "top and bottom", top, bottom) |
|
76
|
|
|
|
|
77
|
|
|
data = pd.DataFrame(data).copy() |
|
78
|
|
|
cols = data.select_dtypes(exclude=["number"]).columns.tolist() |
|
79
|
|
|
data = data[cols] |
|
80
|
|
|
for col in data.columns: |
|
81
|
|
|
if data[col].dtype.name == "category" or data[col].dtype.name == "string": |
|
82
|
|
|
data[col] = data[col].astype("object") |
|
83
|
|
|
|
|
84
|
|
|
if len(cols) == 0: |
|
85
|
|
|
print("No columns with categorical data were detected.") |
|
86
|
|
|
|
|
87
|
|
|
fig = plt.figure(figsize=figsize) |
|
88
|
|
|
gs = fig.add_gridspec(nrows=6, ncols=len(cols), wspace=0.21) |
|
89
|
|
|
|
|
90
|
|
|
for count, col in enumerate(cols): |
|
91
|
|
|
|
|
92
|
|
|
n_unique = data[col].nunique(dropna=True) |
|
93
|
|
|
value_counts = data[col].value_counts() |
|
94
|
|
|
lim_top, lim_bot = top, bottom |
|
95
|
|
|
|
|
96
|
|
|
if n_unique < top + bottom: |
|
97
|
|
|
lim_top = int(n_unique // 2) |
|
98
|
|
|
lim_bot = int(n_unique // 2) + 1 |
|
99
|
|
|
|
|
100
|
|
|
if n_unique <= 2: |
|
101
|
|
|
lim_top = lim_bot = int(n_unique // 2) |
|
102
|
|
|
|
|
103
|
|
|
value_counts_top = value_counts[0:lim_top] |
|
104
|
|
|
value_counts_idx_top = value_counts_top.index.tolist() |
|
105
|
|
|
value_counts_bot = value_counts[-lim_bot:] |
|
106
|
|
|
value_counts_idx_bot = value_counts_bot.index.tolist() |
|
107
|
|
|
|
|
108
|
|
|
if top == 0: |
|
109
|
|
|
value_counts_top = value_counts_idx_top = [] |
|
110
|
|
|
|
|
111
|
|
|
if bottom == 0: |
|
112
|
|
|
value_counts_bot = value_counts_idx_bot = [] |
|
113
|
|
|
|
|
114
|
|
|
data.loc[data[col].isin(value_counts_idx_top), col] = 3 |
|
115
|
|
|
data.loc[data[col].isin(value_counts_idx_bot), col] = -3 |
|
116
|
|
|
data.loc[((data[col] != 3) & (data[col] != -3)), col] = 0 |
|
117
|
|
|
data[col] = data[col].rolling(2, min_periods=1).mean() |
|
118
|
|
|
|
|
119
|
|
|
value_counts_idx_top = [elem[:20] for elem in value_counts_idx_top] |
|
120
|
|
|
value_counts_idx_bot = [elem[:20] for elem in value_counts_idx_bot] |
|
121
|
|
|
|
|
122
|
|
|
# Barcharts |
|
123
|
|
|
ax_top = fig.add_subplot(gs[:1, count : count + 1]) |
|
124
|
|
|
ax_top.bar(value_counts_idx_top, value_counts_top, color=bar_color_top, width=0.85) |
|
125
|
|
|
ax_top.bar(value_counts_idx_bot, value_counts_bot, color=bar_color_bottom, width=0.85) |
|
126
|
|
|
ax_top.set(frame_on=False) |
|
127
|
|
|
ax_top.tick_params(axis="x", labelrotation=90) |
|
128
|
|
|
|
|
129
|
|
|
# Summary stats |
|
130
|
|
|
ax_bottom = fig.add_subplot(gs[1:2, count : count + 1]) |
|
131
|
|
|
plt.subplots_adjust(hspace=0.075) |
|
132
|
|
|
ax_bottom.get_yaxis().set_visible(False) |
|
133
|
|
|
ax_bottom.get_xaxis().set_visible(False) |
|
134
|
|
|
ax_bottom.set(frame_on=False) |
|
135
|
|
|
ax_bottom.text( |
|
136
|
|
|
0, |
|
137
|
|
|
0, |
|
138
|
|
|
f"Unique values: {n_unique}\n\n" |
|
139
|
|
|
f"Top {lim_top} vals: {sum(value_counts_top)} ({sum(value_counts_top)/data.shape[0]*100:.1f}%)\n" |
|
140
|
|
|
f"Bot {lim_bot} vals: {sum(value_counts_bot)} ({sum(value_counts_bot)/data.shape[0]*100:.1f}%)", |
|
141
|
|
|
transform=ax_bottom.transAxes, |
|
142
|
|
|
color="#111111", |
|
143
|
|
|
fontsize=11, |
|
144
|
|
|
) |
|
145
|
|
|
|
|
146
|
|
|
# Heatmap |
|
147
|
|
|
# data = data.astype("int") |
|
148
|
|
|
top_rgb = to_rgb(bar_color_top) |
|
149
|
|
|
color_white = to_rgb("#FFFFFF") |
|
150
|
|
|
bot_rgb = to_rgb(bar_color_bottom) |
|
151
|
|
|
cat_plot_cmap = LinearSegmentedColormap.from_list( |
|
152
|
|
|
"cat_plot_cmap", [bot_rgb, bot_rgb, color_white, top_rgb, top_rgb], N=100 |
|
153
|
|
|
) |
|
154
|
|
|
ax_hm = fig.add_subplot(gs[2:, :]) |
|
155
|
|
|
sns.heatmap(data, cmap=cat_plot_cmap, cbar=False, vmin=-4.25, vmax=4.25, ax=ax_hm) |
|
156
|
|
|
ax_hm.set_yticks(np.round(ax_hm.get_yticks()[0::5], -1)) |
|
157
|
|
|
ax_hm.set_yticklabels(ax_hm.get_yticks()) |
|
158
|
|
|
ax_hm.set_xticklabels( |
|
159
|
|
|
ax_hm.get_xticklabels(), horizontalalignment="center", fontweight="light", fontsize="medium" |
|
160
|
|
|
) |
|
161
|
|
|
ax_hm.tick_params(length=1, colors="#111111") |
|
162
|
|
|
|
|
163
|
|
|
gs.figure.suptitle("Categorical data plot", x=0.5, y=0.91, fontsize=18, color="#111111") |
|
164
|
|
|
|
|
165
|
|
|
return gs |
|
166
|
|
|
|
|
167
|
|
|
|
|
168
|
|
|
# Correlation Matrix |
|
169
|
|
|
def corr_mat( |
|
170
|
|
|
data: pd.DataFrame, |
|
171
|
|
|
split: Optional[str] = None, # Optional[Literal['pos', 'neg', 'high', 'low']] = None, |
|
172
|
|
|
threshold: float = 0, |
|
173
|
|
|
target: Optional[Union[pd.DataFrame, pd.Series, np.ndarray, str]] = None, |
|
174
|
|
|
method: str = "pearson", # Literal['pearson', 'spearman', 'kendall'] = "pearson", |
|
175
|
|
|
colored: bool = True, |
|
176
|
|
|
) -> Union[pd.DataFrame, Any]: |
|
177
|
|
|
""" Returns a color-encoded correlation matrix. |
|
178
|
|
|
|
|
179
|
|
|
Parameters |
|
180
|
|
|
---------- |
|
181
|
|
|
data : pd.DataFrame |
|
182
|
|
|
2D dataset that can be coerced into Pandas DataFrame. If a Pandas DataFrame is provided, the index/column \ |
|
183
|
|
|
information is used to label the plots |
|
184
|
|
|
split : Optional[str], optional |
|
185
|
|
|
Type of split to be performed, by default None |
|
186
|
|
|
{None, 'pos', 'neg', 'high', 'low'} |
|
187
|
|
|
threshold : float, optional |
|
188
|
|
|
Value between 0 <= threshold <= 1, by default 0 |
|
189
|
|
|
target : Optional[Union[pd.DataFrame, str]], optional |
|
190
|
|
|
Specify target for correlation. E.g. label column to generate only the correlations between each feature \ |
|
191
|
|
|
and the label, by default None |
|
192
|
|
|
method : str, optional |
|
193
|
|
|
method: {'pearson', 'spearman', 'kendall'}, by default "pearson" |
|
194
|
|
|
* pearson: measures linear relationships and requires normally distributed and homoscedastic data. |
|
195
|
|
|
* spearman: ranked/ordinal correlation, measures monotonic relationships. |
|
196
|
|
|
* kendall: ranked/ordinal correlation, measures monotonic relationships. Computationally more expensive but \ |
|
197
|
|
|
more robust in smaller dataets than 'spearman' |
|
198
|
|
|
colored : bool, optional |
|
199
|
|
|
If True the negative values in the correlation matrix are colored in red, by default True |
|
200
|
|
|
|
|
201
|
|
|
Returns |
|
202
|
|
|
------- |
|
203
|
|
|
Union[pd.DataFrame, pd.Styler] |
|
204
|
|
|
If colored = True - corr: Pandas Styler object |
|
205
|
|
|
If colored = False - corr: Pandas DataFrame |
|
206
|
|
|
""" |
|
207
|
|
|
|
|
208
|
|
|
# Validate Inputs |
|
209
|
|
|
_validate_input_range(threshold, "threshold", -1, 1) |
|
210
|
|
|
_validate_input_bool(colored, "colored") |
|
211
|
|
|
|
|
212
|
|
|
def color_negative_red(val): |
|
213
|
|
|
color = "#FF3344" if val < 0 else None |
|
214
|
|
|
return "color: %s" % color |
|
215
|
|
|
|
|
216
|
|
|
data = pd.DataFrame(data) |
|
217
|
|
|
|
|
218
|
|
|
if isinstance(target, (str, list, pd.Series, np.ndarray)): |
|
219
|
|
|
target_data = [] |
|
220
|
|
|
if isinstance(target, str): |
|
221
|
|
|
target_data = data[target] |
|
222
|
|
|
data = data.drop(target, axis=1) |
|
223
|
|
|
|
|
224
|
|
|
elif isinstance(target, (list, pd.Series, np.ndarray)): |
|
225
|
|
|
target_data = pd.Series(target) |
|
226
|
|
|
target = target_data.name |
|
227
|
|
|
|
|
228
|
|
|
corr = pd.DataFrame(data.corrwith(target_data, method=method)) |
|
229
|
|
|
corr = corr.sort_values(corr.columns[0], ascending=False) |
|
230
|
|
|
corr.columns = [target] |
|
231
|
|
|
|
|
232
|
|
|
else: |
|
233
|
|
|
corr = data.corr(method=method) |
|
234
|
|
|
|
|
235
|
|
|
corr = _corr_selector(corr, split=split, threshold=threshold) |
|
236
|
|
|
|
|
237
|
|
|
if colored: |
|
238
|
|
|
return corr.style.applymap(color_negative_red).format("{:.2f}", na_rep="-") |
|
239
|
|
|
else: |
|
240
|
|
|
return corr |
|
241
|
|
|
|
|
242
|
|
|
|
|
243
|
|
|
# Correlation matrix / heatmap |
|
244
|
|
|
def corr_plot( |
|
245
|
|
|
data: pd.DataFrame, |
|
246
|
|
|
split: Optional[str] = None, |
|
247
|
|
|
threshold: float = 0, |
|
248
|
|
|
target: Optional[Union[pd.Series, str]] = None, |
|
249
|
|
|
method: str = "pearson", |
|
250
|
|
|
cmap: str = "BrBG", |
|
251
|
|
|
figsize: Tuple = (12, 10), |
|
252
|
|
|
annot: bool = True, |
|
253
|
|
|
dev: bool = False, |
|
254
|
|
|
**kwargs, |
|
255
|
|
|
): |
|
256
|
|
|
""" Two-dimensional visualization of the correlation between feature-columns, excluding NA values. |
|
257
|
|
|
|
|
258
|
|
|
Parameters |
|
259
|
|
|
---------- |
|
260
|
|
|
data : pd.DataFrame |
|
261
|
|
|
2D dataset that can be coerced into Pandas DataFrame. If a Pandas DataFrame is provided, the index/column \ |
|
262
|
|
|
information is used to label the plots |
|
263
|
|
|
split : Optional[str], optional |
|
264
|
|
|
Type of split to be performed {None, 'pos', 'neg', 'high', 'low'}, by default None |
|
265
|
|
|
* None: visualize all correlations between the feature-columns |
|
266
|
|
|
* pos: visualize all positive correlations between the feature-columns above the threshold |
|
267
|
|
|
* neg: visualize all negative correlations between the feature-columns below the threshold |
|
268
|
|
|
* high: visualize all correlations between the feature-columns for which abs(corr) > threshold is True |
|
269
|
|
|
* low: visualize all correlations between the feature-columns for which abs(corr) < threshold is True |
|
270
|
|
|
|
|
271
|
|
|
threshold : float, optional |
|
272
|
|
|
Value between 0 <= threshold <= 1, by default 0 |
|
273
|
|
|
target : Optional[Union[pd.Series, str]], optional |
|
274
|
|
|
Specify target for correlation. E.g. label column to generate only the correlations between each feature \ |
|
275
|
|
|
and the label, by default None |
|
276
|
|
|
method : str, optional |
|
277
|
|
|
method: {'pearson', 'spearman', 'kendall'}, by default "pearson" |
|
278
|
|
|
* pearson: measures linear relationships and requires normally distributed and homoscedastic data. |
|
279
|
|
|
* spearman: ranked/ordinal correlation, measures monotonic relationships. |
|
280
|
|
|
* kendall: ranked/ordinal correlation, measures monotonic relationships. Computationally more expensive \ |
|
281
|
|
|
but more robust in smaller dataets than 'spearman'. |
|
282
|
|
|
|
|
283
|
|
|
cmap : str, optional |
|
284
|
|
|
The mapping from data values to color space, matplotlib colormap name or object, or list of colors, by default \ |
|
285
|
|
|
"BrBG" |
|
286
|
|
|
figsize : Tuple, optional |
|
287
|
|
|
Use to control the figure size, by default (12, 10) |
|
288
|
|
|
annot : bool, optional |
|
289
|
|
|
Use to show or hide annotations, by default True |
|
290
|
|
|
dev : bool, optional |
|
291
|
|
|
Display figure settings in the plot by setting dev = True. If False, the settings are not displayed, by \ |
|
292
|
|
|
default False |
|
293
|
|
|
|
|
294
|
|
|
Keyword Arguments : optional |
|
295
|
|
|
Additional elements to control the visualization of the plot, e.g.: |
|
296
|
|
|
|
|
297
|
|
|
* mask: bool, default True |
|
298
|
|
|
If set to False the entire correlation matrix, including the upper triangle is shown. Set dev = False \ |
|
299
|
|
|
in this case to avoid overlap. |
|
300
|
|
|
* vmax: float, default is calculated from the given correlation coefficients. |
|
301
|
|
|
Value between -1 or vmin <= vmax <= 1, limits the range of the colorbar. |
|
302
|
|
|
* vmin: float, default is calculated from the given correlation coefficients. |
|
303
|
|
|
Value between -1 <= vmin <= 1 or vmax, limits the range of the colorbar. |
|
304
|
|
|
* linewidths: float, default 0.5 |
|
305
|
|
|
Controls the line-width inbetween the squares. |
|
306
|
|
|
* annot_kws: dict, default {'size' : 10} |
|
307
|
|
|
Controls the font size of the annotations. Only available when annot = True. |
|
308
|
|
|
* cbar_kws: dict, default {'shrink': .95, 'aspect': 30} |
|
309
|
|
|
Controls the size of the colorbar. |
|
310
|
|
|
* Many more kwargs are available, i.e. 'alpha' to control blending, or options to adjust labels, ticks ... |
|
311
|
|
|
|
|
312
|
|
|
Kwargs can be supplied through a dictionary of key-value pairs (see above). |
|
313
|
|
|
|
|
314
|
|
|
Returns |
|
315
|
|
|
------- |
|
316
|
|
|
ax: matplotlib Axes |
|
317
|
|
|
Returns the Axes object with the plot for further tweaking. |
|
318
|
|
|
""" |
|
319
|
|
|
|
|
320
|
|
|
# Validate Inputs |
|
321
|
|
|
_validate_input_range(threshold, "threshold", -1, 1) |
|
322
|
|
|
_validate_input_bool(annot, "annot") |
|
323
|
|
|
_validate_input_bool(dev, "dev") |
|
324
|
|
|
|
|
325
|
|
|
data = pd.DataFrame(data) |
|
326
|
|
|
|
|
327
|
|
|
corr = corr_mat(data, split=split, threshold=threshold, target=target, method=method, colored=False) |
|
328
|
|
|
|
|
329
|
|
|
mask = np.zeros_like(corr, dtype=np.bool) |
|
330
|
|
|
|
|
331
|
|
|
if target is None: |
|
332
|
|
|
mask = np.triu(np.ones_like(corr, dtype=np.bool)) |
|
333
|
|
|
|
|
334
|
|
|
vmax = np.round(np.nanmax(corr.where(~mask)) - 0.05, 2) |
|
335
|
|
|
vmin = np.round(np.nanmin(corr.where(~mask)) + 0.05, 2) |
|
336
|
|
|
|
|
337
|
|
|
fig, ax = plt.subplots(figsize=figsize) |
|
338
|
|
|
|
|
339
|
|
|
# Specify kwargs for the heatmap |
|
340
|
|
|
kwargs = { |
|
341
|
|
|
"mask": mask, |
|
342
|
|
|
"cmap": cmap, |
|
343
|
|
|
"annot": annot, |
|
344
|
|
|
"vmax": vmax, |
|
345
|
|
|
"vmin": vmin, |
|
346
|
|
|
"linewidths": 0.5, |
|
347
|
|
|
"annot_kws": {"size": 10}, |
|
348
|
|
|
"cbar_kws": {"shrink": 0.95, "aspect": 30}, |
|
349
|
|
|
**kwargs, |
|
350
|
|
|
} |
|
351
|
|
|
|
|
352
|
|
|
# Draw heatmap with mask and default settings |
|
353
|
|
|
sns.heatmap(corr, center=0, fmt=".2f", **kwargs) |
|
354
|
|
|
|
|
355
|
|
|
ax.set_title(f"Feature-correlation ({method})", fontdict={"fontsize": 18}) |
|
356
|
|
|
|
|
357
|
|
|
# Settings |
|
358
|
|
|
if dev: |
|
359
|
|
|
fig.suptitle( |
|
360
|
|
|
f"\ |
|
361
|
|
|
Settings (dev-mode): \n\ |
|
362
|
|
|
- split-mode: {split} \n\ |
|
363
|
|
|
- threshold: {threshold} \n\ |
|
364
|
|
|
- method: {method} \n\ |
|
365
|
|
|
- annotations: {annot} \n\ |
|
366
|
|
|
- cbar: \n\ |
|
367
|
|
|
- vmax: {vmax} \n\ |
|
368
|
|
|
- vmin: {vmin} \n\ |
|
369
|
|
|
- linewidths: {kwargs['linewidths']} \n\ |
|
370
|
|
|
- annot_kws: {kwargs['annot_kws']} \n\ |
|
371
|
|
|
- cbar_kws: {kwargs['cbar_kws']}", |
|
372
|
|
|
fontsize=12, |
|
373
|
|
|
color="gray", |
|
374
|
|
|
x=0.35, |
|
375
|
|
|
y=0.85, |
|
376
|
|
|
ha="left", |
|
377
|
|
|
) |
|
378
|
|
|
|
|
379
|
|
|
return ax |
|
380
|
|
|
|
|
381
|
|
|
|
|
382
|
|
|
# Distribution plot |
|
383
|
|
|
def dist_plot( |
|
384
|
|
|
data: pd.DataFrame, |
|
385
|
|
|
mean_color: str = "orange", |
|
386
|
|
|
figsize: Tuple = (14, 2), |
|
387
|
|
|
fill_range: Tuple = (0.025, 0.975), |
|
388
|
|
|
hist: bool = False, |
|
389
|
|
|
bins: int = 10, |
|
390
|
|
|
showall: bool = False, |
|
391
|
|
|
kde_kws: Dict[str, Any] = None, |
|
392
|
|
|
rug_kws: Dict[str, Any] = None, |
|
393
|
|
|
fill_kws: Dict[str, Any] = None, |
|
394
|
|
|
font_kws: Dict[str, Any] = None, |
|
395
|
|
|
): |
|
396
|
|
|
""" Two-dimensional visualization of the distribution of numerical features. |
|
397
|
|
|
|
|
398
|
|
|
Parameters |
|
399
|
|
|
---------- |
|
400
|
|
|
data : pd.DataFrame |
|
401
|
|
|
2D dataset that can be coerced into Pandas DataFrame. If a Pandas DataFrame is provided, the index/column \ |
|
402
|
|
|
information is used to label the plots |
|
403
|
|
|
mean_color : str, optional |
|
404
|
|
|
Color of the vertical line indicating the mean of the data, by default "orange" |
|
405
|
|
|
figsize : Tuple, optional |
|
406
|
|
|
Controls the figure size, by default (14, 2) |
|
407
|
|
|
fill_range : Tuple, optional |
|
408
|
|
|
Set the quantiles for shading. Default spans 95% of the data, which is about two std. deviations \ |
|
409
|
|
|
above and below the mean, by default (0.025, 0.975) |
|
410
|
|
|
hist : bool, optional |
|
411
|
|
|
Set to True to display histogram bars in the plot, by default False |
|
412
|
|
|
bins : int, optional |
|
413
|
|
|
Specification of the number of hist bins. Requires hist = True, by default 10 |
|
414
|
|
|
showall : bool, optional |
|
415
|
|
|
Set to True to remove the output limit of 20 plots, by default False |
|
416
|
|
|
kde_kws : Dict[str, Any], optional |
|
417
|
|
|
Keyword arguments for kdeplot(), by default {'color': 'k', 'alpha': 0.7, 'linewidth': 1} |
|
418
|
|
|
rug_kws : Dict[str, Any], optional |
|
419
|
|
|
Keyword arguments for rugplot(), by default {'color': 'brown', 'alpha': 0.5, 'linewidth': 2, 'height': 0.04} |
|
420
|
|
|
fill_kws : Dict[str, Any], optional |
|
421
|
|
|
Keyword arguments to control the fill, by default {'color': 'brown', 'alpha': 0.1} |
|
422
|
|
|
font_kws : Dict[str, Any], optional |
|
423
|
|
|
Keyword arguments to control the font, by default {'color': '#111111', 'weight': 'normal', 'size': 11} |
|
424
|
|
|
|
|
425
|
|
|
Returns |
|
426
|
|
|
------- |
|
427
|
|
|
ax: matplotlib Axes |
|
428
|
|
|
Returns the Axes object with the plot for further tweaking. |
|
429
|
|
|
""" |
|
430
|
|
|
|
|
431
|
|
|
# Validate Inputs |
|
432
|
|
|
_validate_input_range(fill_range[0], "fill_range_lower", 0, 1) |
|
433
|
|
|
_validate_input_range(fill_range[1], "fill_range_upper", 0, 1) |
|
434
|
|
|
_validate_input_smaller(fill_range[0], fill_range[1], "fill_range") |
|
435
|
|
|
_validate_input_bool(hist, "hist") |
|
436
|
|
|
_validate_input_int(bins, "bins") |
|
437
|
|
|
_validate_input_range(bins, "bins", 0, data.shape[0]) |
|
438
|
|
|
_validate_input_bool(showall, "showall") |
|
439
|
|
|
|
|
440
|
|
|
# Handle dictionary defaults |
|
441
|
|
|
kde_kws = {"alpha": 0.7, "linewidth": 1.5} if kde_kws is None else kde_kws.copy() |
|
442
|
|
|
rug_kws = ( |
|
443
|
|
|
{"color": "brown", "alpha": 0.5, "linewidth": 2, "height": 0.04} |
|
444
|
|
|
if rug_kws is None |
|
445
|
|
|
else rug_kws.copy() |
|
446
|
|
|
) |
|
447
|
|
|
fill_kws = {"color": "brown", "alpha": 0.1} if fill_kws is None else fill_kws.copy() |
|
448
|
|
|
font_kws = {"color": "#111111", "weight": "normal", "size": 11} if font_kws is None else font_kws.copy() |
|
449
|
|
|
|
|
450
|
|
|
data = pd.DataFrame(data.copy()).dropna(axis=1, how="all") |
|
451
|
|
|
cols = list(data.select_dtypes(include=["number"]).columns) |
|
452
|
|
|
data = data[cols] |
|
453
|
|
|
|
|
454
|
|
|
if len(cols) == 0: |
|
455
|
|
|
print("No columns with numeric data were detected.") |
|
456
|
|
|
return |
|
457
|
|
|
|
|
458
|
|
|
elif len(cols) >= 20 and showall is False: |
|
459
|
|
|
print( |
|
460
|
|
|
f"Note: The number of numerical features is very large ({len(cols)}), please consider splitting the data. " |
|
461
|
|
|
"Showing plots for the first 20 numerical features. Override this by setting showall=True." |
|
462
|
|
|
) |
|
463
|
|
|
cols = cols[:20] |
|
464
|
|
|
|
|
465
|
|
|
for col in cols: |
|
466
|
|
|
dropped_values = data[col].isna().sum() |
|
467
|
|
|
if dropped_values > 0: |
|
468
|
|
|
col_data = data[col].dropna(axis=0) |
|
469
|
|
|
print(f"Dropped {dropped_values} missing values from column {col}.") |
|
470
|
|
|
|
|
471
|
|
|
else: |
|
472
|
|
|
col_data = data[col] |
|
473
|
|
|
|
|
474
|
|
|
_, ax = plt.subplots(figsize=figsize) |
|
475
|
|
|
ax = sns.distplot( |
|
476
|
|
|
col_data, |
|
477
|
|
|
bins=bins, |
|
478
|
|
|
hist=hist, |
|
479
|
|
|
rug=True, |
|
480
|
|
|
kde_kws=kde_kws, |
|
481
|
|
|
rug_kws=rug_kws, |
|
482
|
|
|
hist_kws={"alpha": 0.5, "histtype": "step"}, |
|
483
|
|
|
) |
|
484
|
|
|
|
|
485
|
|
|
# Vertical lines and fill |
|
486
|
|
|
x, y = ax.lines[0].get_xydata().T |
|
487
|
|
|
ax.fill_between( |
|
488
|
|
|
x, |
|
489
|
|
|
y, |
|
490
|
|
|
where=((x >= np.quantile(col_data, fill_range[0])) & (x <= np.quantile(col_data, fill_range[1]))), |
|
491
|
|
|
label=f"{fill_range[0]*100:.1f}% - {fill_range[1]*100:.1f}%", |
|
492
|
|
|
**fill_kws, |
|
493
|
|
|
) |
|
494
|
|
|
|
|
495
|
|
|
mean = np.mean(col_data) |
|
496
|
|
|
std = scipy.stats.tstd(col_data) |
|
497
|
|
|
ax.vlines( |
|
498
|
|
|
x=mean, ymin=0, ymax=np.interp(mean, x, y), ls="dotted", color=mean_color, lw=2, label="mean" |
|
499
|
|
|
) |
|
500
|
|
|
ax.vlines( |
|
501
|
|
|
x=np.median(col_data), |
|
502
|
|
|
ymin=0, |
|
503
|
|
|
ymax=np.interp(np.median(col_data), x, y), |
|
504
|
|
|
ls=":", |
|
505
|
|
|
color=".3", |
|
506
|
|
|
label="median", |
|
507
|
|
|
) |
|
508
|
|
|
ax.vlines( |
|
509
|
|
|
x=[mean - std, mean + std], |
|
510
|
|
|
ymin=0, |
|
511
|
|
|
ymax=[np.interp(mean - std, x, y), np.interp(mean + std, x, y)], |
|
512
|
|
|
ls=":", |
|
513
|
|
|
color=".5", |
|
514
|
|
|
label="\u03BC \u00B1 \u03C3", |
|
515
|
|
|
) |
|
516
|
|
|
|
|
517
|
|
|
ax.set_ylim(0) |
|
518
|
|
|
ax.set_xlim(ax.get_xlim()[0] * 1.15, ax.get_xlim()[1] * 1.15) |
|
519
|
|
|
|
|
520
|
|
|
# Annotations and legend |
|
521
|
|
|
ax.text(0.01, 0.85, f"Mean: {np.round(mean,2)}", fontdict=font_kws, transform=ax.transAxes) |
|
522
|
|
|
ax.text(0.01, 0.7, f"Std. dev: {np.round(std,2)}", fontdict=font_kws, transform=ax.transAxes) |
|
523
|
|
|
ax.text( |
|
524
|
|
|
0.01, |
|
525
|
|
|
0.55, |
|
526
|
|
|
f"Skew: {np.round(scipy.stats.skew(col_data),2)}", |
|
527
|
|
|
fontdict=font_kws, |
|
528
|
|
|
transform=ax.transAxes, |
|
529
|
|
|
) |
|
530
|
|
|
ax.text( |
|
531
|
|
|
0.01, |
|
532
|
|
|
0.4, |
|
533
|
|
|
f"Kurtosis: {np.round(scipy.stats.kurtosis(col_data),2)}", # Excess Kurtosis |
|
534
|
|
|
fontdict=font_kws, |
|
535
|
|
|
transform=ax.transAxes, |
|
536
|
|
|
) |
|
537
|
|
|
ax.text(0.01, 0.25, f"Count: {np.round(len(col_data))}", fontdict=font_kws, transform=ax.transAxes) |
|
538
|
|
|
ax.legend(loc="upper right") |
|
539
|
|
|
|
|
540
|
|
|
return ax |
|
541
|
|
|
|
|
542
|
|
|
|
|
543
|
|
|
# Missing value plot |
|
544
|
|
|
def missingval_plot( |
|
545
|
|
|
data: pd.DataFrame, |
|
546
|
|
|
cmap: str = "PuBuGn", |
|
547
|
|
|
figsize: Tuple = (20, 20), |
|
548
|
|
|
sort: bool = False, |
|
549
|
|
|
spine_color: str = "#EEEEEE", |
|
550
|
|
|
): |
|
551
|
|
|
""" Two-dimensional visualization of the missing values in a dataset. |
|
552
|
|
|
|
|
553
|
|
|
Parameters |
|
554
|
|
|
---------- |
|
555
|
|
|
data : pd.DataFrame |
|
556
|
|
|
2D dataset that can be coerced into Pandas DataFrame. If a Pandas DataFrame is provided, the index/column \ |
|
557
|
|
|
information is used to label the plots |
|
558
|
|
|
cmap : str, optional |
|
559
|
|
|
Any valid colormap can be used. E.g. 'Greys', 'RdPu'. More information can be found in the matplotlib \ |
|
560
|
|
|
documentation, by default "PuBuGn" |
|
561
|
|
|
figsize : Tuple, optional |
|
562
|
|
|
Use to control the figure size, by default (20, 20) |
|
563
|
|
|
sort : bool, optional |
|
564
|
|
|
Sort columns based on missing values in descending order and drop columns without any missing values, \ |
|
565
|
|
|
by default False |
|
566
|
|
|
spine_color : str, optional |
|
567
|
|
|
Set to 'None' to hide the spines on all plots or use any valid matplotlib color argument, by default "#EEEEEE" |
|
568
|
|
|
|
|
569
|
|
|
Returns |
|
570
|
|
|
------- |
|
571
|
|
|
GridSpec |
|
572
|
|
|
gs: Figure with array of Axes objects |
|
573
|
|
|
""" |
|
574
|
|
|
|
|
575
|
|
|
# Validate Inputs |
|
576
|
|
|
_validate_input_bool(sort, "sort") |
|
577
|
|
|
|
|
578
|
|
|
data = pd.DataFrame(data) |
|
579
|
|
|
|
|
580
|
|
|
if sort: |
|
581
|
|
|
mv_cols_sorted = data.isna().sum(axis=0).sort_values(ascending=False) |
|
582
|
|
|
final_cols = ( |
|
583
|
|
|
mv_cols_sorted.drop(mv_cols_sorted[mv_cols_sorted.values == 0].keys().tolist()).keys().tolist() |
|
584
|
|
|
) |
|
585
|
|
|
data = data[final_cols] |
|
586
|
|
|
print("Displaying only columns with missing values.") |
|
587
|
|
|
|
|
588
|
|
|
# Identify missing values |
|
589
|
|
|
mv_total, mv_rows, mv_cols, _, mv_cols_ratio = _missing_vals(data).values() |
|
590
|
|
|
total_datapoints = data.shape[0] * data.shape[1] |
|
591
|
|
|
|
|
592
|
|
|
if mv_total == 0: |
|
593
|
|
|
print("No missing values found in the dataset.") |
|
594
|
|
|
else: |
|
595
|
|
|
# Create figure and axes |
|
596
|
|
|
fig = plt.figure(figsize=figsize) |
|
597
|
|
|
gs = fig.add_gridspec(nrows=6, ncols=6, left=0.1, wspace=0.05) |
|
598
|
|
|
ax1 = fig.add_subplot(gs[:1, :5]) |
|
599
|
|
|
ax2 = fig.add_subplot(gs[1:, :5]) |
|
600
|
|
|
ax3 = fig.add_subplot(gs[:1, 5:]) |
|
601
|
|
|
ax4 = fig.add_subplot(gs[1:, 5:]) |
|
602
|
|
|
|
|
603
|
|
|
# ax1 - Barplot |
|
604
|
|
|
colors = plt.get_cmap(cmap)(mv_cols / np.max(mv_cols)) # color bars by height |
|
605
|
|
|
ax1.bar(range(len(mv_cols)), np.round((mv_cols_ratio) * 100, 2), color=colors) |
|
606
|
|
|
ax1.get_xaxis().set_visible(False) |
|
607
|
|
|
ax1.set(frame_on=False, xlim=(-0.5, len(mv_cols) - 0.5)) |
|
608
|
|
|
ax1.set_ylim(0, np.max(mv_cols_ratio) * 100) |
|
609
|
|
|
ax1.grid(linestyle=":", linewidth=1) |
|
610
|
|
|
ax1.yaxis.set_major_formatter(ticker.PercentFormatter(decimals=0)) |
|
611
|
|
|
ax1.tick_params(axis="y", colors="#111111", length=1) |
|
612
|
|
|
|
|
613
|
|
|
# annotate values on top of the bars |
|
614
|
|
|
for rect, label in zip(ax1.patches, mv_cols): |
|
615
|
|
|
height = rect.get_height() |
|
616
|
|
|
ax1.text( |
|
617
|
|
|
0.1 + rect.get_x() + rect.get_width() / 2, |
|
618
|
|
|
height + 0.5, |
|
619
|
|
|
label, |
|
620
|
|
|
ha="center", |
|
621
|
|
|
va="bottom", |
|
622
|
|
|
rotation="90", |
|
623
|
|
|
alpha=0.5, |
|
624
|
|
|
fontsize="11", |
|
625
|
|
|
) |
|
626
|
|
|
|
|
627
|
|
|
ax1.set_frame_on(True) |
|
628
|
|
|
for _, spine in ax1.spines.items(): |
|
629
|
|
|
spine.set_visible(True) |
|
630
|
|
|
spine.set_color(spine_color) |
|
631
|
|
|
ax1.spines["top"].set_color(None) |
|
632
|
|
|
|
|
633
|
|
|
# ax2 - Heatmap |
|
634
|
|
|
sns.heatmap(data.isna(), cbar=False, cmap="binary", ax=ax2) |
|
635
|
|
|
ax2.set_yticks(np.round(ax2.get_yticks()[0::5], -1)) |
|
636
|
|
|
ax2.set_yticklabels(ax2.get_yticks()) |
|
637
|
|
|
ax2.set_xticklabels( |
|
638
|
|
|
ax2.get_xticklabels(), horizontalalignment="center", fontweight="light", fontsize="12" |
|
639
|
|
|
) |
|
640
|
|
|
ax2.tick_params(length=1, colors="#111111") |
|
641
|
|
|
for _, spine in ax2.spines.items(): |
|
642
|
|
|
spine.set_visible(True) |
|
643
|
|
|
spine.set_color(spine_color) |
|
644
|
|
|
|
|
645
|
|
|
# ax3 - Summary |
|
646
|
|
|
fontax3 = {"color": "#111111", "weight": "normal", "size": 14} |
|
647
|
|
|
ax3.get_xaxis().set_visible(False) |
|
648
|
|
|
ax3.get_yaxis().set_visible(False) |
|
649
|
|
|
ax3.set(frame_on=False) |
|
650
|
|
|
|
|
651
|
|
|
ax3.text( |
|
652
|
|
|
0.025, |
|
653
|
|
|
0.875, |
|
654
|
|
|
f"Total: {np.round(total_datapoints/1000,1)}K", |
|
655
|
|
|
transform=ax3.transAxes, |
|
656
|
|
|
fontdict=fontax3, |
|
657
|
|
|
) |
|
658
|
|
|
ax3.text( |
|
659
|
|
|
0.025, 0.675, f"Missing: {np.round(mv_total/1000,1)}K", transform=ax3.transAxes, fontdict=fontax3 |
|
660
|
|
|
) |
|
661
|
|
|
ax3.text( |
|
662
|
|
|
0.025, |
|
663
|
|
|
0.475, |
|
664
|
|
|
f"Relative: {np.round(mv_total/total_datapoints*100,1)}%", |
|
665
|
|
|
transform=ax3.transAxes, |
|
666
|
|
|
fontdict=fontax3, |
|
667
|
|
|
) |
|
668
|
|
|
ax3.text( |
|
669
|
|
|
0.025, |
|
670
|
|
|
0.275, |
|
671
|
|
|
f"Max-col: {np.round(mv_cols.max()/data.shape[0]*100)}%", |
|
672
|
|
|
transform=ax3.transAxes, |
|
673
|
|
|
fontdict=fontax3, |
|
674
|
|
|
) |
|
675
|
|
|
ax3.text( |
|
676
|
|
|
0.025, |
|
677
|
|
|
0.075, |
|
678
|
|
|
f"Max-row: {np.round(mv_rows.max()/data.shape[1]*100)}%", |
|
679
|
|
|
transform=ax3.transAxes, |
|
680
|
|
|
fontdict=fontax3, |
|
681
|
|
|
) |
|
682
|
|
|
|
|
683
|
|
|
# ax4 - Scatter plot |
|
684
|
|
|
ax4.get_yaxis().set_visible(False) |
|
685
|
|
|
for _, spine in ax4.spines.items(): |
|
686
|
|
|
spine.set_color(spine_color) |
|
687
|
|
|
ax4.tick_params(axis="x", colors="#111111", length=1) |
|
688
|
|
|
|
|
689
|
|
|
ax4.scatter(mv_rows, range(len(mv_rows)), s=mv_rows, c=mv_rows, cmap=cmap, marker=".", vmin=1) |
|
690
|
|
|
ax4.set_ylim((0, len(mv_rows))[::-1]) # limit and invert y-axis |
|
691
|
|
|
ax4.set_xlim(0, max(mv_rows) + 0.5) |
|
692
|
|
|
ax4.grid(linestyle=":", linewidth=1) |
|
693
|
|
|
|
|
694
|
|
|
gs.figure.suptitle("Missing value plot", x=0.45, y=0.94, fontsize=18, color="#111111") |
|
695
|
|
|
|
|
696
|
|
|
return gs |
|
697
|
|
|
|
akanz1 /
klib
