如何检测大Pandas的上升和下降趋势？

Question

对于用下面的代码生成的绘图，我想通过pandas逻辑生成一个信号。在

当曲线比上一个局部最小值高出+3个点（或更多）时，输出信号应从-4变为-2。当曲线比上一个局部最大值低2个点（或更少）时，它应该从-2变回-4。在

图1显示了由下面代码生成的曲线。图2大致显示了输出信号的外观。在

地块1：

地块2：

代码：

import matplotlib

matplotlib.use('QT5Agg')
import matplotlib.pyplot as plt
import numpy as np

a = np.arange(5)
b = np.arange(5, -4, -1)
c = np.arange(-4, 7, .5)
d = np.arange(7, 2, -1)
e = np.arange(2, 6, .2)
f = np.arange(6, -3, -1)
g = np.arange(-3, 2, .25)

r1 = np.append(a, b)
r2 = np.append(r1, c)
r3 = np.append(r2, d)
r4 = np.append(r3, e)
r5 = np.append(r4, f)
r6 = np.append(r5, g)

plt.rcParams['font.size'] = 6

fig, ax1 = plt.subplots()
ax1.plot(r6,'g-o',markersize=3)



plt.annotate('start upward', xy=(0,0), textcoords='data',)
plt.annotate('end upward', xy=(3,3), textcoords='data',)

plt.annotate('start downward', xy=(5,5), textcoords='data',)
plt.annotate('end downward', xy=(7,3), textcoords='data',)

plt.annotate('start upward', xy=(14,-4), textcoords='data',)
plt.annotate('end upward', xy=(20,-1), textcoords='data',)

plt.annotate('start downward', xy=(36,7), textcoords='data',)
plt.annotate('end downward', xy=(38,5), textcoords='data',)

plt.annotate('start upward', xy=(41,2), textcoords='data',)
plt.annotate('end upward', xy=(56,5), textcoords='data',)

plt.annotate('start downward', xy=(61,6), textcoords='data',)
plt.annotate('end downward', xy=(63,4), textcoords='data',)

plt.annotate('start upward', xy=(70,-3), textcoords='data',)
plt.annotate('end upward', xy=(82,0), textcoords='data',)

ax1.minorticks_on()
ax1.grid(b=True, which='major', color='g', linestyle='-')
ax1.grid(b=True, which='minor', color='y', linestyle='--')
plt.show()

Answer 1

IIUC：

s = pd.Series([0, 5, 0, -4, -1, 2, 4, 7, 2, 3, 4, 5, 6, -3, -2, -1, 0, 1, 2])

s.plot()
ax = s.diff().ge(0).mul(1).plot(drawstyle='steps', c='r', secondary_y=True)
ax.set_ylim(0, 8)

Answer 2

我想你想要这个：

s = pd.Series(np.concatenate((a,b,c,d,e,f,g,)))

# is increasing
incr = s.diff().ge(0)

# shifted trend (local minima)
shifted = incr.ne(incr.shift())

# local max
local_max = shifted & (~incr)


# thresholding function
def thresh(x, threshold=3, step=2):
    ret = pd.Series([0]*len(x), index=x.index)
    t = x.min() + threshold
    ret.loc[x.gt(t)] = step
    return ret

signal = s.groupby(local_max.cumsum()).apply(thresh)
signal += s.min()

# draw
fig, ax = plt.subplots(figsize=(10,6))
s.plot(ax=ax)
signal.plot(drawstyle='steps', ax=ax)
plt.show()

输出：