python matplotlib 画棘状图和叠加柱状图

2022年1月2日更新：

# 绘制棘状图的函数
def jizhuangtu(b, colors=['#1f77b4','lightseagreen'], labels=None, title=None, text_on=True):     # 绘制棘状图的函数
    '''
    b：Serie, 由groupby(x).y.values()得到
    colors:list, 分类的颜色（y轴分类）
    labels:list, 分类的标签（y轴分类）
    title：标签的名称/plt.legend的参数
    text_on：是否添加文字标签
    '''
    fig, ax = plt.subplots()

    for n, value in enumerate(b.index.levels[0]):
        c = b[value].sort_index()
        c = c/c.sum()
        globals()['width%s'%(n+1)] = b[value].sum()/b.sum()   # 计算样本量占比 width1:男 width2:女
        globals()['x%s'%(n+1)] = 0 if n==0 else globals()['x%s'%n]+globals()['width%s'%(n)]/2+0.02+globals()['width%s'%(n+1)]/2 # 两个柱状图保留0.1的间隙
        plt.bar(globals()['x%s'%(n+1)], c.cumsum()[::-1], color=colors, width=globals()['width%s'%(n+1)])
        if text_on:         # 如果text_on=True, 那么添加文字标签
            for h, s in zip(c[::-1], c):
                plt.text(globals()['x%s'%(n+1)], h, format(s, '.2%') ,horizontalalignment='center', verticalalignment='bottom')

    plt.ylim(0,1)
    plt.xlim(0-width1/2, globals()['x%s'%(n+1)]+globals()['width%s'%(n+1)]/2)  # xn+widthn/2
#     xticks = [i +'({:.2%})'.format(globals()['width%s'%(n+1)]) for n, i in enumerate(b.index.levels[0])]   # 给x标签添加样本占比，添加占比太占位置了，且观看者可能难以理解
    plt.xticks([globals()['x%s'%(i)] for i in range(1, n+2)], b.index.levels[0])

    # 添加图例
    labels = b.index.levels[1] if labels==None else labels
    [plt.bar(0,0,color=colors[n], label=lab) for n,lab in enumerate(labels)]
    num1 = 1.01; num2 = 0.5; num3 = 3; num4 = 0
    plt.legend(bbox_to_anchor=(num1, num2), title=title, loc=num3, borderaxespad=num4)
    
    # 去掉边框
    for loc in ['top', 'left', 'bottom', 'right']:
        ax.spines[loc].set_visible(False)

jizhuang_data

jizhuangtu(jizhuang_data, labels=['逾期', '未逾期'])

---

之前的版本：

bar_data = pht.groupby('性别').是否为前线城市.value_counts().unstack().apply(lambda x: x/x.sum(), axis=1)
bar_data.plot(kind='bar', stacked=True)

bar_data

---

以前的版本：

写了一个可以画棘状图的函数

def jizhuangtu(b, colors=['#bebebe','#ffd700'], labels=None):     # 绘制棘状图的函数
    fig, ax = plt.subplots()

    for n, value in enumerate(b.index.levels[0]):
        c = b[value].sort_index()
        c = c/c.sum()
        globals()['width%s'%(n+1)] = b[value].sum()/b.sum()   # 计算样本量占比 width1:男 width2:女
        globals()['x%s'%(n+1)] = 0 if n==0 else globals()['x%s'%n]+globals()['width%s'%(n)]/2+0.02+globals()['width%s'%(n+1)]/2 # 两个柱状图保留0.1的间隙
        plt.bar(globals()['x%s'%(n+1)], c.cumsum()[::-1], color=colors, width=globals()['width%s'%(n+1)])

    plt.ylim(0,1)
    plt.xlim(0-width1/2, globals()['x%s'%(n+1)]+globals()['width%s'%(n+1)]/2)  # xn+widthn/2
    plt.xticks([globals()['x%s'%(i)] for i in range(1, n+2)], b.index.levels[0])

    # 添加图例
    labels = b.index.levels[1] if labels==None else labels
    [plt.bar(0,0,color=colors[n],label=lab) for n,lab in enumerate(labels)]
    num1 = 1.01; num2 = 0.5; num3 = 3; num4 = 0
    plt.legend(bbox_to_anchor=(num1, num2), loc=num3, borderaxespad=num4)
    # 去掉边框
    for loc in ['top', 'left', 'bottom', 'right']:
        ax.spines[loc].set_visible(False)

# 生成数据
by = 20; bn = 80
gy = 2; gn = 18
a = pd.DataFrame(columns=['性别','是否出险'])
a = a.append([dict(性别='男',是否出险=1)]*by, ignore_index=True)
a = a.append([dict(性别='男',是否出险=0)]*bn, ignore_index=True)
a = a.append([dict(性别='女',是否出险=1)]*gy, ignore_index=True)
a = a.append([dict(性别='女',是否出险=0)]*gn, ignore_index=True)
a = a.append([dict(性别='未知',是否出险=0)]*20, ignore_index=True)

b = a.groupby('性别').是否出险.value_counts()
b

jizhuangtu(b)

修改图例

jizhuangtu(b, labels=['a','b'])

注：记得将索引补全

b
>>> 性别  是否出险	a
	女		0	18
	女		1	2
	未知	0	20
	男		0	80
	男		1	20

想要补全 (未知，1)

full_index = pd.MultiIndex.from_product([['男','女','未知'],[0,1]])
b = b.reindex(full_index)
b
>>>
	男   0    80.0
	     1    20.0
	女   0    18.0
	     1     2.0
	未知  0    20.0
	     1     NaN

画累积/叠加柱状图 https://zhuanlan.zhihu.com/p/71840687

Survived_m = train.Survived[train.Sex =='male'].value_counts()
Survived_f = train.Survived[train.Sex =='female'].value_counts()

df = pd.DataFrame({u'男性':Survived_m,u'女性':Survived_f})
df.plot(kind='bar',stacked =True)
plt.title(u'按性别看获救情况')
plt.xlabel(u'性别')
plt.ylabel(u'人数')

优化上述代码(一行解决）

trainData.groupby('p1_gender').y1_is_purchase.value_counts().unstack().plot(kind='bar', stacked=True)