您也可以编写自己的Python C扩展。

这当然不是最简单的方法，但与使用np相比，这将使您运行得更快，内存效率更高。堆积:作为建筑块的堆积

// moving_average.c
#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
#include <Python.h>
#include <numpy/arrayobject.h>

static PyObject *moving_average(PyObject *self, PyObject *args) {
    PyObject *input;
    int64_t window_size;
    PyArg_ParseTuple(args, "Ol", &input, &window_size);
    if (PyErr_Occurred()) return NULL;
    if (!PyArray_Check(input) || !PyArray_ISNUMBER((PyArrayObject *)input)) {
        PyErr_SetString(PyExc_TypeError, "First argument must be a numpy array with numeric dtype");
        return NULL;
    }
    
    int64_t input_size = PyObject_Size(input);
    double *input_data;
    if (PyArray_AsCArray(&input, &input_data, (npy_intp[]){ [0] = input_size }, 1, PyArray_DescrFromType(NPY_DOUBLE)) != 0) {
        PyErr_SetString(PyExc_TypeError, "Failed to simulate C array of type double");
        return NULL;
    }
    
    int64_t output_size = input_size - window_size + 1;
    PyObject *output = PyArray_SimpleNew(1, (npy_intp[]){ [0] = output_size }, NPY_DOUBLE);
    double *output_data = PyArray_DATA((PyArrayObject *)output);
    
    double cumsum_before = 0;
    double cumsum_after = 0;
    for (int i = 0; i < window_size; ++i) {
        cumsum_after += input_data[i];
    }
    for (int i = 0; i < output_size - 1; ++i) {
        output_data[i] = (cumsum_after - cumsum_before) / window_size;
        cumsum_after += input_data[i + window_size];
        cumsum_before += input_data[i];
    }
    output_data[output_size - 1] = (cumsum_after - cumsum_before) / window_size;

    return output;
}

static PyMethodDef methods[] = {
    {
        "moving_average", 
        moving_average, 
        METH_VARARGS, 
        "Rolling mean of numpy array with specified window size"
    },
    {NULL, NULL, 0, NULL}
};

static struct PyModuleDef moduledef = {
    PyModuleDef_HEAD_INIT,
    "moving_average",
    "C extension for finding the rolling mean of a numpy array",
    -1,
    methods
};

PyMODINIT_FUNC PyInit_moving_average(void) {
    PyObject *module = PyModule_Create(&moduledef);
    import_array();
    return module;
}

METH_VARARGS specifies that the method only takes positional arguments. PyArg_ParseTuple allows you to parse these positional arguments. By using PyErr_SetString and returning NULL from the method, you can signal that an exception has occurred to the Python interpreter from the C extension. PyArray_AsCArray allows your method to be polymorphic when it comes to input array dtype, alignment, whether the array is C-contiguous (See "Can a numpy 1d array not be contiguous?") etc. without needing to create a copy of the array. If you instead used PyArray_DATA, you'd need to deal with this yourself. PyArray_SimpleNew allows you to create a new numpy array. This is similar to using np.empty. The array will not be initialized, and might contain non-deterministic junk which could surprise you if you forget to overwrite it.

构建C扩展

# setup.py
from setuptools import setup, Extension
import numpy

setup(
  ext_modules=[
    Extension(
      'moving_average',
      ['moving_average.c'],
      include_dirs=[numpy.get_include()]
    )
  ]
)

# python setup.py build_ext --build-lib=.

基准

import numpy as np

# Our compiled C extension:
from moving_average import moving_average as moving_average_c

# Answer by Jaime using npcumsum
def moving_average_cumsum(a, n) :
    ret = np.cumsum(a, dtype=float)
    ret[n:] = ret[n:] - ret[:-n]
    return ret[n - 1:] / n

# Answer by yatu using np.convolve
def moving_average_convolve(a, n):
    return np.convolve(a, np.ones(n), 'valid') / n

a = np.random.rand(1_000_000)
print('window_size = 3')
%timeit moving_average_c(a, 3)
%timeit moving_average_cumsum(a, 3)
%timeit moving_average_convolve(a, 3)

print('\nwindow_size = 100')
%timeit moving_average_c(a, 100)
%timeit moving_average_cumsum(a, 100)
%timeit moving_average_convolve(a, 100)

window_size = 3
958 µs ± 4.68 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)
4.52 ms ± 15.4 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
809 µs ± 463 ns per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

window_size = 100
977 µs ± 937 ns per loop (mean ± std. dev. of 7 runs, 1,000 loops each)
6.16 ms ± 19.1 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
14.2 ms ± 12.4 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

从Numpy 1.20开始，sliding_window_view提供了一种在元素窗口中滑动/滚动的方法。然后你可以分别取平均值。

例如，对于一个4元素的窗口:

from numpy.lib.stride_tricks import sliding_window_view

# values = np.array([5, 3, 8, 10, 2, 1, 5, 1, 0, 2])
np.average(sliding_window_view(values, window_shape = 4), axis=1)
# array([6.5, 5.75, 5.25, 4.5, 2.25, 1.75, 2])

注意sliding_window_view的中间结果:

# values = np.array([5, 3, 8, 10, 2, 1, 5, 1, 0, 2])
sliding_window_view(values, window_shape = 4)
# array([[ 5,  3,  8, 10],
#        [ 3,  8, 10,  2],
#        [ 8, 10,  2,  1],
#        [10,  2,  1,  5],
#        [ 2,  1,  5,  1],
#        [ 1,  5,  1,  0],
#        [ 5,  1,  0,  2]])

NumPy缺乏特定领域的函数可能是由于核心团队的纪律和对NumPy主要指令的忠实:提供n维数组类型，以及用于创建和索引这些数组的函数。像许多基本目标一样，这个目标并不小，NumPy出色地完成了它。

更大的SciPy包含更大的特定于领域的库集合(被SciPy开发人员称为子包)——例如，数值优化(optimize)、信号处理(signal)和积分(integrate)。

我的猜测是，您要查找的函数至少在SciPy子包中的一个(SciPy。也许信号);然而，我将首先在SciPy scikit集合中查找，确定相关的scikit并在其中寻找感兴趣的函数。

Scikits是基于NumPy/SciPy独立开发的包，并针对特定的技术规程(例如，Scikits -image, Scikits -learn等)，其中几个(特别是用于数值优化的令人钦佩的OpenOpt)在选择位于相对较新的Scikits主题之下很久以前就得到了高度重视，成熟的项目。Scikits主页上列出了大约30个这样的Scikits，尽管其中至少有几个已经不再处于积极的开发中。

按照这个建议，你会发现scikits-timeseries;但是，该软件包已不再处于积极开发阶段;实际上，Pandas已经成为AFAIK，事实上的基于numpy的时间序列库。

Pandas有几个函数可以用来计算移动平均线;其中最简单的可能是rolling_mean，你可以这样使用:

>>> # the recommended syntax to import pandas
>>> import pandas as PD
>>> import numpy as NP

>>> # prepare some fake data:
>>> # the date-time indices:
>>> t = PD.date_range('1/1/2010', '12/31/2012', freq='D')

>>> # the data:
>>> x = NP.arange(0, t.shape[0])

>>> # combine the data & index into a Pandas 'Series' object
>>> D = PD.Series(x, t)

现在，只需调用函数rolling_mean，传入Series对象和窗口大小，在下面的例子中是10天。

>>> d_mva = PD.rolling_mean(D, 10)

>>> # d_mva is the same size as the original Series
>>> d_mva.shape
    (1096,)

>>> # though obviously the first w values are NaN where w is the window size
>>> d_mva[:3]
    2010-01-01         NaN
    2010-01-02         NaN
    2010-01-03         NaN

验证它是否有效。，将原系列中的值10 - 15与用滚动平均值平滑的新系列进行比较

>>> D[10:15]
     2010-01-11    2.041076
     2010-01-12    2.041076
     2010-01-13    2.720585
     2010-01-14    2.720585
     2010-01-15    3.656987
     Freq: D

>>> d_mva[10:20]
      2010-01-11    3.131125
      2010-01-12    3.035232
      2010-01-13    2.923144
      2010-01-14    2.811055
      2010-01-15    2.785824
      Freq: D

The function rolling_mean, along with about a dozen or so other function are informally grouped in the Pandas documentation under the rubric moving window functions; a second, related group of functions in Pandas is referred to as exponentially-weighted functions (e.g., ewma, which calculates exponentially moving weighted average). The fact that this second group is not included in the first (moving window functions) is perhaps because the exponentially-weighted transforms don't rely on a fixed-length window

实现这一点的一个简单方法是使用np.卷积。 这背后的思想是利用离散卷积的计算方式，并使用它来返回滚动平均值。这可以通过与np序列进行卷积来实现。长度等于我们想要的滑动窗口长度。

为了做到这一点，我们可以定义以下函数:

def moving_average(x, w):
    return np.convolve(x, np.ones(w), 'valid') / w

该函数将对序列x和长度为w的序列进行卷积。注意，所选模式是有效的，因此卷积积只对序列完全重叠的点给出。

一些例子:

x = np.array([5,3,8,10,2,1,5,1,0,2])

对于窗口长度为2的移动平均线，我们有:

moving_average(x, 2)
# array([4. , 5.5, 9. , 6. , 1.5, 3. , 3. , 0.5, 1. ])

对于长度为4的窗口:

moving_average(x, 4)
# array([6.5 , 5.75, 5.25, 4.5 , 2.25, 1.75, 2.  ])

卷积是怎么工作的?

让我们更深入地看看离散卷积是如何计算的。 下面的函数旨在复制np。卷积计算输出值:

def mov_avg(x, w):
    for m in range(len(x)-(w-1)):
        yield sum(np.ones(w) * x[m:m+w]) / w 

对于上面的同一个例子，也会得到:

list(mov_avg(x, 2))
# [4.0, 5.5, 9.0, 6.0, 1.5, 3.0, 3.0, 0.5, 1.0]

所以每一步要做的就是求1数组和当前窗口之间的内积。在这种情况下，乘以np.ones(w)是多余的，因为我们直接取序列的和。

下面是一个计算第一个输出的例子，这样会更清楚一些。假设我们想要一个w=4的窗口:

[1,1,1,1]
[5,3,8,10,2,1,5,1,0,2]
= (1*5 + 1*3 + 1*8 + 1*10) / w = 6.5

下面的输出将被计算为:

  [1,1,1,1]
[5,3,8,10,2,1,5,1,0,2]
= (1*3 + 1*8 + 1*10 + 1*2) / w = 5.75

依此类推，在所有重叠完成后返回序列的移动平均值。

下面是一个使用numba的快速实现(注意类型)。注意它确实包含移位的nan。

import numpy as np
import numba as nb

@nb.jit(nb.float64[:](nb.float64[:],nb.int64),
        fastmath=True,nopython=True)
def moving_average( array, window ):    
    ret = np.cumsum(array)
    ret[window:] = ret[window:] - ret[:-window]
    ma = ret[window - 1:] / window
    n = np.empty(window-1); n.fill(np.nan)
    return np.concatenate((n.ravel(), ma.ravel())) 

我要么使用公认答案的解决方案，稍微修改以使输出和输入的长度相同，要么使用另一个答案的评论中提到的熊猫版本。我在这里用一个可重复的例子来总结两者，以供将来参考:

import numpy as np
import pandas as pd

def moving_average(a, n):
    ret = np.cumsum(a, dtype=float)
    ret[n:] = ret[n:] - ret[:-n]
    return ret / n

def moving_average_centered(a, n):
    return pd.Series(a).rolling(window=n, center=True).mean().to_numpy()

A = [0, 0, 1, 2, 4, 5, 4]
print(moving_average(A, 3))    
# [0.         0.         0.33333333 1.         2.33333333 3.66666667 4.33333333]
print(moving_average_centered(A, 3))
# [nan        0.33333333 1.         2.33333333 3.66666667 4.33333333 nan       ]