python - 如何 reshape 多个时间序列信号以与 sns.tsplot 一起使用？

Question

我正在尝试 reshape 数据，如下所示:

   t  y0  y1  y2
0  0  -1   0   1
1  1   0   1   2
2  2   1   2   3
3  3   2   3   4
4  4   3   4   5

变成这样的东西:

    t  trial signal  value
 0  0  0     y       -1
 1  0  1     y        0
 2  0  2     y        1
 3  1  0     y        0
 4  1  1     y        1
 5  1  2     y        2
 6  2  0     y        1
 7  2  1     y        2
 8  2  2     y        3
 9  3  0     y        2
10  3  1     y        3
11  3  2     y        4
12  4  0     y        3
13  4  1     y        4
14  4  2     y        5

这样我就可以将其输入到 sns.tsplot 中。

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns

fig = plt.figure()
num_points = 5

# Create some dummy line signals and assemble a data frame
t = np.arange(num_points)
y0 = t - 1
y1 = t
y2 = t + 1
df = pd.DataFrame(np.vstack((t, y0, y1, y2)).transpose(), columns=['t', 'y0', 'y1', 'y2'])
print(df)

# Do some magic transformations
df = pd.melt(df, id_vars=['t'])
print(df)

# Plot the time-series data
sns.tsplot(time="t", value="value", unit="trial", condition="signal", data=df, ci=[68, 95])

plt.savefig("dummy.png")
plt.close()

我希望能够实现这一点:

https://stanford.edu/~mwaskom/software/seaborn/generated/seaborn.tsplot.html http://pandas.pydata.org/pandas-docs/stable/reshaping.html

Answer 1

我认为你可以使用melt对于 reshape ，通过 indexing with str 获取第一个和第二个字符最后sort_values重新排序列:

df1 = pd.melt(df, id_vars=['t'])
#create helper Series
variable = df1['variable']
#extract second char, convert to int
df1['trial'] = variable.str[1].astype(int)
#extract first char
df1['signal'] = variable.str[0]
#sort values by column t, reset index
df1 = df1.sort_values('t').reset_index(drop=True)
#reorder columns
df1 = df1[['t','trial','signal','value']]
print df1
    t trial signal  value
0   0     0      y     -1
1   0     1      y      0
2   0     2      y      1
3   1     0      y      0
4   1     1      y      1
5   1     2      y      2
6   2     0      y      1
7   2     1      y      2
8   2     2      y      3
9   3     0      y      2
10  3     1      y      3
11  3     2      y      4
12  4     0      y      3
13  4     1      y      4
14  4     2      y      5

另一种解决方案，如果 signal 列中的所有值都仅为 y:

#remove y from column name, first value of column names is same
df.columns = df.columns[:1].tolist() + [int(col[1]) for col in df.columns[1:]]
print df
   t  0  1  2
0  0 -1  0  1
1  1  0  1  2
2  2  1  2  3
3  3  2  3  4
4  4  3  4  5

df1 = pd.melt(df, id_vars=['t'], var_name=['trial'])
#all values in column signal are y
df1['signal'] = 't'
#sort values by column t, reset index
df1 = df1.sort_values('t').reset_index(drop=True)
#reorder columns
df1 = df1[['t','trial','signal','value']]
print df1
    t trial signal  value
0   0     0      t     -1
1   0     1      t      0
2   0     2      t      1
3   1     0      t      0
4   1     1      t      1
5   1     2      t      2
6   2     0      t      1
7   2     1      t      2
8   2     2      t      3
9   3     0      t      2
10  3     1      t      3
11  3     2      t      4
12  4     0      t      3
13  4     1      t      4
14  4     2      t      5