python - 序列索引图 python

Question

我正在尝试绘制一组时间排序值的序列，其中每个值都由不同的颜色表示。例如，如果我的值(value)观是这样的

state_seq = [2, 2, 2, 2, 0, 0, 0, 1, 3, 6, 3, 3, 3, 0, 0]
time_seq = ['2013-05-29 09:31:00', '2013-05-29 09:46:00', '2013-07-23 12:28:00', '2013-07-23 01:53:00', '2013-08-05 02:02:00', '2013-08-05 02:08:00', '2013-08-05 04:28:00', '2013-08-06 10:20:00', '2013-08-06 03:03:00', '2013-08-06 04:13:00', '2013-08-06 04:17:00', '2013-08-06 04:36:00', '2013-08-07 11:07:00', '2013-08-07 12:28:00', '2013-08-07 12:31:00']

我希望 state_seq 的每个唯一元素用颜色表示并绘制序列。截至目前，我正在使用seaborn palplot 以一种简单的方式完成此任务，例如，

color_set = ["#95a5a6", "#34495e","#9b59b6", "#3498db", "#800000", "#2ecc71", 'y', '#FF4500']
palette = list()
for s in state_seq:
  palette.append(color_set[s])    

sns.palplot(palette)

这给了我这样的东西(这个输出可能与代码片段不完全匹配 - 我编辑了代码以获得更好的清晰度)

这种方法有一个缺点，就是我无法在 x 轴上表示我的时间标签。是否有更好的 python 替代方案，可能类似于此处解释的 R 包 TraMineR Is it possible to make a graph with pattern fills using TraMineR and R base graphs?

Answer 1

我假设您希望每个值都有方框，而不管每个值之间的时间间隔如何。要获得指示时间间隙的宽度，您可以 convert time_seq to datetime objects然后直接用 matplotlib 绘制日期。

import matplotlib.pyplot as plt
import numpy as np

state_seq = [2, 2, 2, 2, 0, 0, 0, 1, 3, 6, 3, 3, 3, 0, 0]
time_seq = ['2013-05-29 09:31:00', '2013-05-29 09:46:00', '2013-07-23 12:28:00', '2013-07-23 01:53:00', '2013-08-05 02:02:00', '2013-08-05 02:08:00', '2013-08-05 04:28:00', '2013-08-06 10:20:00', '2013-08-06 03:03:00', '2013-08-06 04:13:00', '2013-08-06 04:17:00', '2013-08-06 04:36:00', '2013-08-07 11:07:00', '2013-08-07 12:28:00', '2013-08-07 12:31:00']
color_set = ["#95a5a6", "#34495e","#9b59b6", "#3498db", "#800000", "#2ecc71", 'y', '#FF4500']

for i in range(len(time_seq)):
    # fill the yaxis with our colour for each time_seq entry for width 1
    plt.fill_between((i,i+1), 1, color=color_set[state_seq[i]])

# keep the coloured boxes square
plt.axis("scaled")

plt.xlim(0, len(time_seq))
plt.axes().get_yaxis().set_visible(False)

# add custom tick labels based on time_seq and set to middle of boxes
plt.xticks(np.arange(0.5,len(time_seq)+0.5, 1),  time_seq, rotation='vertical')
# remove xaxis ticks
plt.axes().get_xaxis().set_ticks_position("none")

plt.savefig("boxes.png", bbox_inches="tight")

生产: