如何根据分仓的共同条件来识别分仓事件？(scipy.binned_statistic)

Question

我使用scipy.binned_statistic来获取一个bin内的点的频率，这样。

h, xedge, yedge, binindex = scipy.stats.binned_statistic_2d(X, Y, Y, statistic='mean', bins=160)

我可以用下面的方法过滤掉某些bin。

filter = list(np.argwhere(h > 5).flatten())

由此我可以从我感兴趣的数据的边缘和yedges中得到bin edgescenter.

从这些感兴趣的bin中得到原始数据的最python方式是什么？例如，我如何获得包含在超过5个点的bin中的原始数据？

Answer 1

是的，通过一些索引魔法是可以的。我不确定这是否是最Pythonic的方法，但应该很接近。

1d的解决方案是使用 stats.binned_statistic:

from scipy import stats
import numpy as np

values = np.array([1.0, 1.0, 2.0, 1.5, 3.0]) # not used with 'count'
x = np.array([1, 1, 1, 4, 7, 7, 7])

statistic, bin_edges, binnumber = stats.binned_statistic(x, values, 'count', bins=3)
print(statistic)
print(bin_edges)
print(binnumber)

# find the bins with equal or more than three events
# if you are using custom bins where events can be lower or
# higher than your specified bins -> handle this

# get the bin numbers according to some condition
idx_bin = np.where(statistic >= 3)[0]
print(idx_bin)

# A binnumber of i means the corresponding value is 
# between (bin_edges[i-1], bin_edges[i]).
# -> increment the bin indices by one 

idx_bin += 1
print(idx_bin)

# the rest is easy, get the boolean mask and apply it
is_event = np.in1d(binnumber, idx_bin)
events = x[is_event]
print(events)

对于2d或nd，你可以多次使用上面的解决方案，并将其结合起来。is_event 每个维度的掩码使用 np.logical_and (2d)或 np.logical_and.reduce((x, y, z)) (nd，见 此处).

2d的解决方法是用 stats.binned_statistic_2d 基本上是一样的。

from scipy import stats
import numpy as np

x = np.array([1, 1.5, 2.0, 4, 5.5, 1.5, 7, 1])
y = np.array([1.0, 7.0, 1.0, 3, 7, 7, 7, 1])
values = np.ones_like(x) # not used with 'count'

# check keyword expand_binnumbers, use non-linearized
# as they can be used as indices without flattening
ret = stats.binned_statistic_2d(x, 
                                y,
                                values,
                                'count',
                                bins=2,
                                expand_binnumbers=True)

print(ret.statistic)
print('binnumber', ret.binnumber)

binnumber = ret.binnumber
statistic = ret.statistic
# find the bins with equal or more than three events
# if you are using custom bins where events can be lower or
# higher than your specified bins -> handle this

# get the bin numbers according to some condition
idx_bin_x, idx_bin_y = np.where(statistic >= 3)#[0]
print(idx_bin_x)
print(idx_bin_y)

# A binnumber of i means the corresponding value is 
# between (bin_edges[i-1], bin_edges[i]).
# -> increment the bin indices by one 
idx_bin_x += 1
idx_bin_y += 1

print(idx_bin_x)
print(idx_bin_y)

# the rest is easy, get the boolean mask and apply it
is_event_x = np.in1d(binnumber[0], idx_bin_x)
is_event_y = np.in1d(binnumber[1], idx_bin_y)
is_event_xy = np.logical_and(is_event_x, is_event_y)

events_x = x[is_event_xy]
events_y = y[is_event_xy]

print('x', events_x)
print('y', events_y)