explore.py 文件源码

def corr_fit_plot(s, k):
    """
    ?????????lag=k?
    ???
    """
    n = len(s)
    x = []; y = []
    for i in range(0,n-k):
        x.append([s[i]])
        y.append([s[i+k]])
    plt.scatter(x,y)

#    # using sklearn
#    re = LinearRegression()
#    re.fit(x,y)
#    pred = re.predict(x)
#    coef = re.coef_
#    plt.plot(x,pred,'r-')

    # least square by myself
    x = np.array(x)
    y = np.array(y)
    one = np.ones((x.shape[0],1))
    x = np.concatenate((one,np.array(x)),axis=1)
    coefs = np.dot(pinv(x),y)
    pred = coefs[0]*x[:,0] + coefs[1]*x[:,1]
    coef = coefs[1]
    plt.plot(x[:,1],pred,'r-')

    plt.title('Corr=%s'%coef+' Lag=%s'%k)
    plt.show()    

    return coef

#corr_fit_plot(s_power_consumption.values,1)
#corr_fit_plot(s_power_consumption.values,3)
#corr_fit_plot(s_power_consumption.values,5)
#corr_fit_plot(s_power_consumption.values,7)