Building Machine Learning Systems with Python 1

 1 import scipy as sp
 2 data = sp.genfromtxt("web_traffic.tsv", delimiter="\t")
 3 x = data[:,0]
 4 y = data[:,1]
 5 x = x[~sp.isnan(y)]
 6 y = y[~sp.isnan(y)]
 7 import matplotlib.pyplot as plt
 8 plt.scatter(x,y)
 9 plt.title("Web traffic over the last month")
10 plt.xlabel("Time")
11 plt.ylabel("Hits/hour")
12 plt.xticks([w*7*24 for w in range(10)],
13            ['week %i'%w for w in range(10)])
14 
15 def error(f,x,y):
16     return sp.sum((f(x)-y)**2)
17 
18 fp1, residuals, rank, sv, rcomd = sp.polyfit(x,y,1,full=True)
19 f1 = sp.poly1d(fp1)
20 fx = sp.linspace(0,x[-1], 1000)
21 plt.plot(fx, f1(fx), linewidth=4)
22 plt.legend(["d=%i" % f1.order], loc="upper left")
23 
24 f2p = sp.polyfit(x, y, 2)
25 f2 = sp.poly1d(f2p)
26 fx = sp.linspace(0,x[-1], 1000)
27 plt.plot(fx, f2(fx), linewidth=4)
28 plt.legend(["d=%i" % f2.order], loc="upper left")
29 
30 f3p = sp.polyfit(x, y, 3)
31 f3 = sp.poly1d(f3p)
32 fx = sp.linspace(0,x[-1], 1000)
33 plt.plot(fx, f3(fx), linewidth=4)
34 plt.legend(["d=%i" % f3.order], loc="upper left")
35 
36 plt.autoscale(tight=True)
37 plt.grid()
38 plt.show()

Note:

1>polyfit()  在图中画出要拟合的线。给定数据x，y 以及期望的多项式阶，它可以找到一个模型,并且返回函数模型的参数。

2>poly1d() 根据参数创建模型函数。