通过时间戳绘制numpy数组的直方图

时间:2014-10-06 18:39:05

标签: python numpy matplotlib scipy

我的数组有一对unix时间戳和值。

[[  1.40170249e+09   9.00000000e+01]
 [  1.40170249e+09   9.10000000e+01]
 [  1.40170249e+09   9.20000000e+01]
 ..., 
 [  1.41149703e+09   1.09000000e+02]
 [  1.41149703e+09   1.06000000e+02]
 [  1.41149703e+09   1.06000000e+02]]

我设法用pyplot.hist(array[:,1]); pyplot.show()绘制整个第二列的直方图。但我真正想做的是,按日{bin array[:,1](由数组[:,0]中的unix时间戳导出),并将它们绘制为堆叠直方图,每个(彩色)堆栈代表一个天。什么是最好的方法呢?

1 个答案:

答案 0 :(得分:1)

由于您有groupby参与其中,因此使用pandas是有意义的:

In [192]:
import pandas as pd
import numpy as np
import time
A = np.array([[  1.40170249e+09,   9.00000000e+01],
             [  1.40170249e+09,   9.10000000e+01],
             [  1.40170249e+09,   9.20000000e+01],
             [  1.41149703e+09,   1.09000000e+02],
             [  1.41149703e+09,   1.06000000e+02],
             [  1.41149703e+09,   1.06000000e+02]])
df = pd.DataFrame(A, columns=['date', 'val'])
df['date'] = df.date.map(lambda x: time.gmtime(x))
print df
                                   date  val
0    (2014, 6, 2, 9, 48, 10, 0, 153, 0)   90
1    (2014, 6, 2, 9, 48, 10, 0, 153, 0)   91
2    (2014, 6, 2, 9, 48, 10, 0, 153, 0)   92
3  (2014, 9, 23, 18, 30, 30, 1, 266, 0)  109
4  (2014, 9, 23, 18, 30, 30, 1, 266, 0)  106
5  (2014, 9, 23, 18, 30, 30, 1, 266, 0)  106
In [193]:

grp_obj = df.groupby(df.date.map(lambda x: time.strftime('%Y-%m-%d', x)))
plt.hist([value.val.values for grp, value in grp_obj],
         stacked=True, 
         label=[grp for grp, value in grp_obj])
plt.legend()
Out[193]:
<matplotlib.legend.Legend at 0x10902d950>

enter image description here

此外,您还需要按年度分组,以避免将不同月份/年份的日子组合在一起。