Question

我使用未经校准的指南针在现场收集了简单的调查数据。实现这个问题，在现场，轴承比较了一个好的指南针和未校准的指南针和11个轴承记录的差异。该图显示差异非常接近sin函数。我希望将多项式（阶数3）拟合到这个结果函数，以使用未校准的罗盘校正调查数据。我的曲线拟合程序产生不良拟合曲线。任何人都可以看到有什么问题吗？

import numpy as np
import scipy
import pylab
correctCompass=\
np.array([134.4,112.6,069.7,051.1,352.5,314.6,218.3,258.2,237.8,186.5,153.7])
errorCompass=\
np.array([131.6,108.9,065.6,047.0,349.8,314.0,284.6,262.7,243.4,189.8,153.2])
# sort compass values
for i in range(0,11):
   for j in range(i+1,11):
       if correctCompass[i] > correctCompass[j]:
          tmp=correctCompass[j]
          correctCompass[j]=correctCompass[i]
          correctCompass[i]=tmp
          tmp=errorCompass[j]
          errorCompass[j]=errorCompass[i]
          errorCompass[i]=tmp

diff = correctCompass - errorCompass + 15.0
height=diff.max() + 16.0
polycoeffs = scipy.polyfit(correctCompass, diff, 3)

# fit the data with a polynomial
yfit = scipy.polyval(polycoeffs,correctCompass)

pylab.plot(correctCompass, diff, 'k.')
pylab.plot(correctCompass, yfit, 'r-')

pylab.axis([0,360,-10.0,height])
pylab.show()

Answer 1

polyfit工作正常，问题是diff中的一个负面因素降低了适合度，但是当您将yaxis的最低值设置为-10时，未在图中显示p>

diff = array([ 19.1,  19.1,  18.7,  17.8,  15.5,  11.7, -51.3,   9.4,  10.5, 15.6,  17.7])

如果您发表评论pylab.axis([0,360,-10.0,height])，您会看到“问题”

此外，您可以使用以下三行代替两个嵌套的for循环来改进并使代码更具可读性：

sort = np.argsort(correctCompass)
correctCompass = correctCompass[sort]
errorCompass = errorCompass[sort]

Scipy Bodgy曲线适合

1 个答案: