Pythonで集計した棒グラフを出力したい

前提・実現したいこと

下記のようなpythonのデータフレーム型でa列が正解ラベル（1.0が正解、0.0が不正解）、b列が予測値のデータとします。
各予測値の範囲における正解率の棒グラフを作成したいのですが、集計方法で苦戦してます。

import numpy as np
import pandas as pd

df = pd.DataFrame(data= [[0.0,0.728463888168335,0.0]
,[1.0,0.027108728885650635,430]
,[1.0,0.5831845998764038,280]
,[0.0,0.03684014081954956,0.0]
,[1.0,0.9906975030899048,100]
,[0.0,0.035909026861190796,0.0]
,[1.0,0.8771868944168091,120]
,[0.0,0.18819177150726318,0.0]
,[0.0,0.7758017778396606,0.0]
,[0.0,0.13000166416168213,0.0]
,[1.0,0.23878684639930725,110]
,[0.0,0.02938440442085266,0.0]
,[0.0,0.13557732105255127,0.0]
,[0.0,0.07250338792800903,0.0]
,[0.0,0.5754911303520203,0.0]
,[0.0,0.3734626770019531,0.0]
,[0.0,0.5296110510826111,0.0]
,[0.0,0.53096604347229,0.0]
,[0.0,0.0026207268238067627,0.0]
,[1.0,0.9533392190933228,560]], columns=['a', 'b', 'c'])

作成したい棒グラフのイメージは下記のような形です。
予測値の範囲で正解ラベルが1.0の割合を表現したいです。

試したこと

Pandasのgroupbyを使用してデータの集計を試みましたが、うまくいきませんでした。

行動規範の内容に同意します

回答3件

こんな感じで良いのではないでしょうか

Python
1import numpy as np
2import pandas as pd
3import matplotlib.pyplot as plt
4
5df = pd.DataFrame(data= [[0.0,0.728463888168335,0.0]
6,[1.0,0.027108728885650635,430]
7,[1.0,0.5831845998764038,280]
8,[0.0,0.03684014081954956,0.0]
9,[1.0,0.9906975030899048,100]
10,[0.0,0.035909026861190796,0.0]
11,[1.0,0.8771868944168091,120]
12,[0.0,0.18819177150726318,0.0]
13,[0.0,0.7758017778396606,0.0]
14,[0.0,0.13000166416168213,0.0]
15,[1.0,0.23878684639930725,110]
16,[0.0,0.02938440442085266,0.0]
17,[0.0,0.13557732105255127,0.0]
18,[0.0,0.07250338792800903,0.0]
19,[0.0,0.5754911303520203,0.0]
20,[0.0,0.3734626770019531,0.0]
21,[0.0,0.5296110510826111,0.0]
22,[0.0,0.53096604347229,0.0]
23,[0.0,0.0026207268238067627,0.0]
24,[1.0,0.9533392190933228,560]], columns=['a', 'b', 'c'])
25
26df['範囲'] = pd.cut(df['b'],
27                    bins=np.arange(0,1.1,0.1),
28                    labels=[f'0.{v}-0.{v+1}' for v in range(0,10)])
29df.groupby('範囲')['a'].mean().plot.bar()
30df.groupby(pd.cut(df['b'],np.arange(0,1.1,0.1)))['a'].mean().plot.bar()
31
32plt.show()

やっていることは b列を pandas.cut() にて binning し、

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.cut.html

範囲ごとに'a'列の値の平均値を groupby().mean() にて求めてPlotしているだけです。

ラベルの表記が少し変わりますが、

Python
1df.groupby(pd.cut(df['b'],np.arange(0,1.1,0.1)))['a'].mean().plot.bar()

でワンライナーでも書けますね（蛇足）。

投稿2020/02/12 02:03

magichan

総合スコア15898

補足として

Python
1bins = np.arange(0, 1.1, 0.1)
2
3b_cut = np.floor(df['b'].to_numpy() * 10).astype(int)
4# b_cut = np.searchsorted(bins[1:], df['b'].to_numpy(), 'right') でも可
5result = np.bincount(b_cut, df['a'].to_numpy()) / np.bincount(b_cut)
6
7result
8# array([0.16666667, 0.        , 1.        , 0.        ,        nan,
9#        0.25      ,        nan, 0.        , 1.        , 1.        ])
10
11labels = bins[:-1]
12plt.bar(labels, result, width=0.05)

投稿2020/02/12 06:26

編集2020/02/12 06:36

kirara0048

総合スコア1399

少し雑ですが、こういう形で大丈夫でしょうか。ビン分割を使えばいいと思います。

python
1import numpy as np
2import pandas as pd
3
4df = pd.DataFrame(data= [[0.0,0.728463888168335,0.0]
5,[1.0,0.027108728885650635,430]
6,[1.0,0.5831845998764038,280]
7,[0.0,0.03684014081954956,0.0]
8,[1.0,0.9906975030899048,100]
9,[0.0,0.035909026861190796,0.0]
10,[1.0,0.8771868944168091,120]
11,[0.0,0.18819177150726318,0.0]
12,[0.0,0.7758017778396606,0.0]
13,[0.0,0.13000166416168213,0.0]
14,[1.0,0.23878684639930725,110]
15,[0.0,0.02938440442085266,0.0]
16,[0.0,0.13557732105255127,0.0]
17,[0.0,0.07250338792800903,0.0]
18,[0.0,0.5754911303520203,0.0]
19,[0.0,0.3734626770019531,0.0]
20,[0.0,0.5296110510826111,0.0]
21,[0.0,0.53096604347229,0.0]
22,[0.0,0.0026207268238067627,0.0]
23,[1.0,0.9533392190933228,560]], columns=['a', 'b', 'c'])
24
25res = df.groupby(['b','a']).count().reset_index()
26res['d']=pd.cut(res['b'], [0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0])
27
28result = res.groupby(['d','a']).count()
29result = result.fillna(0)
30
31def transformation_sample(s):
32    return (s / s.sum() * 100)
33
34result = result.groupby(['d']).transform(transformation_sample)
35result = result.reset_index()
36result
37

この状態で、resultにはcの列に割合の情報が入っています。b,c列の情報が必要な場合は、色々やり方がありますが事前に取り出して繋げるなどのやり方があります。

	d	a	b	c
0	(0.0, 0.1]	0.0	83.333333	83.333333
1	(0.0, 0.1]	1.0	16.666667	16.666667
2	(0.1, 0.2]	0.0	100.000000	100.000000
3	(0.1, 0.2]	1.0	0.000000	0.000000
4	(0.2, 0.3]	0.0	0.000000	0.000000
5	(0.2, 0.3]	1.0	100.000000	100.000000

で、プロットします。

Python
1import matplotlib.pyplot as plt
2%matplotlib inline
3
4ax = result[result.a==1.0].plot(x="d", y="c",kind='bar',color='b')
5ax.legend(['割合'],fontsize=14,loc='upper left', bbox_to_anchor=(1, 1), frameon=False)
6ax.set_xlabel("",fontsize=14)
7labels = ax.get_xticklabels()
8ax.set_xticklabels(labels, rotation=45)