python - 在数据框中使用分类变量，沿着散点图中的线对线图进行着色

Question

我正在计算每秒传感器数据的“状态/事件”变量(字符串)。有 12 个州，数据平均跨度为 10-12 天。我正在构建一个带有传感器数据参数的每秒日志查看器，以及“状态/事件”。绘图如下例所示。我正在尝试根据 "category" 变量的值对 "battle_deaths" 列进行着色。绘图中有一个 color 属性，但在我见过的所有示例中，它采用数值，我无法将分类值“映射”颜色。请参阅下面的当前输出和预期输出( overdraw 输出)

#dataframe with time index
data = {
        'date': ['2014-05-01 18:47:05.069722', '2014-05-01 18:47:06.119994',
 '2014-05-01 18:47:07.178768', '2014-05-01 18:47:08.230071', 
'2014-05-01 18:47:09.230071', '2014-05-01 18:47:10.280592', 
'2014-05-01 18:47:11.332662', '2014-05-01 18:47:12.385109', 
'2014-05-01 18:47:13.436523', '2014-05-01 18:47:14.486877'], 
        'battle_deaths': [34, 25, 26, 15, 15, 14, 26, 25, 62, 41],
        'category' : ["A", "A","A","A","C","A","B","C","B","B"],
        'chicken_dinners':["4000", "5000", "6000", "-1000","4500", 
                            "5900", "6300", "6712","7788","4681"]
       }

df = pd.DataFrame(data, columns = ['date', 'battle_deaths', 'category', 'chicken_dinners'])
df['date'] = pd.to_datetime(df['date'])
df.index = df['date']
del df['date']

print(df)   

>     date  battle_deaths   category    
>     2014-05-01 18:47:05.069722    34  A
>     2014-05-01 18:47:06.119994    25  A
>     2014-05-01 18:47:07.178768    26  A
>     2014-05-01 18:47:08.230071    15  A
>     2014-05-01 18:47:09.230071    15  C
>     2014-05-01 18:47:10.280592    14  A
>     2014-05-01 18:47:11.332662    26  B
>     2014-05-01 18:47:12.385109    25  C
>     2014-05-01 18:47:13.436523    62  B
>     2014-05-01 18:47:14.486877    41  B



#plot code
random_x = df.index

traceC1 = go.Scattergl(
    x=random_x,
    y=df["battle_deaths"],
    mode='lines+ markers',
    name="battle_deaths ",
    hoverinfo='x'
)
traceC2 = go.Scattergl(
    x=random_x,
    y=df["chicken_dinners"],
    mode='lines',
    name="chicken_dinners",
    hoverinfo='y'
)  

#append traces to the above colored plot, no need to color other plots
fig_circ = tools.make_subplots(rows=2, cols=1, shared_xaxes=True)
fig_circ.append_trace(traceC1, 1, 1)
fig_circ.append_trace(traceC2, 2, 1)

#custom scales on different sensor data channels
#scaling is important and can't autoscale, because data has 'spikes' all over the place

fig_circ['layout'].update(
                            height=1000, width=1600,
                            margin = dict(l = 100, r =0, t=0, b= 0),
                            yaxis = dict(range = [0, 100],
                            yaxis2 = dict(range = [-50, 500])
                        )

plotly.offline.plot(fig_circ, filename='sample.html')

Answer 1

目前(Feb 2019)没有简单/直接的方法可以做到这一点。

一种可能的解决方案是:

• 用不同颜色绘制多条迹线
• 通过legendgroup对相同的颜色进行分组
• 如果类别已绘制，则将 showlegend 设置为 False

下面的代码可以使用一些优化，但它应该可以帮助您入门。

import pandas as pd
import plotly
plotly.offline.init_notebook_mode()

# taken from the original question
data = {
        'date': ['2014-05-01 18:47:05.069722', '2014-05-01 18:47:06.119994',
'2014-05-01 18:47:07.178768', '2014-05-01 18:47:08.230071', 
'2014-05-01 18:47:09.230071', '2014-05-01 18:47:10.280592', 
'2014-05-01 18:47:11.332662', '2014-05-01 18:47:12.385109', 
'2014-05-01 18:47:13.436523', '2014-05-01 18:47:14.486877'], 
        'battle_deaths': [34, 25, 26, 15, 15, 14, 26, 25, 62, 41],
        'category' : ["A", "A","A","A","C","A","B","C","B","B"]
       }

df = pd.DataFrame(data, columns = ['date', 'battle_deaths', 'category'])
df['date'] = pd.to_datetime(df['date'])
df.index = df['date']
del df['date']

# just an empty figure
fig = plotly.graph_objs.Figure()

# a dict which maps your categorical values to colors
colors = {'A': 'orange',
          'B': 'green',
          'C': 'red'}

# the list which stores categories which were already plotted
already_plotted = []

for i in range(df.shape[0] + 1):
    # create a new trace if the category changes or at the end of the data frame
    if i in (0, df.shape[0]) or cat != df.iloc[i, ]['category']:
        if i != 0:
            if i != df.shape[0]:
                x.append(df.iloc[i,].name)
                y.append(df.iloc[i,]['battle_deaths'])
            trace = plotly.graph_objs.Scatter(x=x, y=y, 
                                              legendgroup=cat,  # group identical categories
                                              showlegend=cat not in already_plotted,  # hide legend if already plotted
                                              name=cat,
                                              marker={'color': colors[df.iloc[i - 1, ]['category']]})
            fig.add_trace(trace)
            already_plotted.append(cat)

        if i == df.shape[0]:
            continue
        cat = df.iloc[i, ]['category']
        x = []
        y = []    

    x.append(df.iloc[i,].name)
    y.append(df.iloc[i,]['battle_deaths'])

plotly.offline.iplot(fig)