Extrapolation from data plotted using matplotlib

I have 10 values of x and y in my file.

Is there any way that I can extrapolate the graph ie make it into a continous function and increasing its range for other x-values in matplotlib ??

I would even be thankful if anyone can tell 开发者_如何转开发me if there is any other software that I can use. I basically want that these 10 values get approximated to a continous function so that I can know the y-value at some random x point.

below i use Scipy, but the same functions (polyval and polyfit) are also in NumPy; NumPy is a Matplotlib dependency so you can import those two functions from there if you don't have SciPy installed.

import numpy as NP
from scipy import polyval, polyfit
from matplotlib import pyplot as PLT

n=10   # 10 data points
# make up some data
x = NP.linspace(0, 1, n)
y = 7*x**2 - 5*x + 3
# add some noise 
noise = NP.random.normal(.5, .3, 10)
y += noise

# the shape of the data suggests a 2d polynomial, so begin there
# a, b, c are the polynomial coefficients: ax^2 + bx + c
a, b, c = polyfit(x, y, 2)
y_pred = polyval([a, b, c], x)    # y_pred refers to predicted values of y

# how good is the fit?
# calculate MSE:
MSE = NP.sqrt( NP.sum((y_pred-y)**2)/10 )
# MSE = .2

# now use the model polynomial to generate y values based on x values outside 
# the range of the original data:
x_out = NP.linspace(0, 2, 20)   # choose 20 points, 10 in, 10 outside original range
y_pred = polyval([a, b, c], x_out)

# now plot the original data points and the polynomial fit through them
fig = PLT.figure()
ax1 = fig.add_subplot(111)

ax1.plot(x, y, 'g.', x_out, y_pred, 'b-' )

PLT.show()

If you are using SciPy (Scientific Python) you can try scipy.interp1d. See the manual for an example.

Otherwise, any decent spreadsheet software should be able to do spline interpolation and give you a nice smooth graph.

Beware of extrapolation, though. If you don't have a good model for your data you might get completely unrelated data when extrapolating outside your input range.

Example (EDIT):

from scipy.interpolate import interp1d

# the available data points
x = [1, 2, 3]
y = [10, 20, 30]

# return a function f, such that f(x) is the interpolated value at 'x'
f = interp1d(x, y, kind='cubic')

You can now compute the function f(x) at any point x. For example print f(2.5) will return the interpolated value for x=2.5.

Most of what you want can be found here: http://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html

But don't extrapolate, at least until you are absolutely sure that you know what you are doing.