def test_lwlr(self):
dataMat, labelMat = lr.loadDataSet("ex0.txt")
# 对dataMat第一行的样本数据进行预测
w = lwlr.lwlr(dataMat[0], dataMat, labelMat, 1)
print("\n w == %s" % (w))
yHat = lwlr.lwlrTest(dataMat, dataMat, labelMat, 1)
print("\n yHat == %s" % (yHat))
def test_corrcoef(self):
dataMat, labelMat = lr.loadDataSet("ex0.txt")
w = lr.standRegres(dataMat, labelMat)
print("\n w == %s" % (w))
xMat = mat(dataMat)
yMat = mat(labelMat)
# 预测值
yHat = xMat * w
# 计算预估值与真实值的相关系数
n = corrcoef(yHat.T, yMat)
print("\n corrcoef == %s" % (n))
def test_abalone(self):
data_set, label_set = lr.loadDataSet("abalone.txt")
train_data_set = data_set[0:99]
train_label_set = label_set[0:99]
yHat01 = lwlr.lwlrTest(train_data_set, train_data_set, train_label_set,
0.1)
yHat1 = lwlr.lwlrTest(train_data_set, train_data_set, train_label_set,
1)
yHat10 = lwlr.lwlrTest(train_data_set, train_data_set, train_label_set,
10)
# 计算不同k值的误差值：
error01 = abalone.rssError(train_label_set, yHat01.T)
# error01 == 56.7987246777
print("\n error01 == %s" % (error01))
error1 = abalone.rssError(train_label_set, yHat1.T)
# error1 == 429.89056187
print("\n error1 == %s" % (error1))
error10 = abalone.rssError(train_label_set, yHat10.T)
# error10 == 549.118170883
print("\n error10 == %s" % (error10))
# 由上述结果发现，使用最小的核 k=0.1 得到最低的误差。但是对于新数据则不然。
new_data_set = data_set[100:199]
new_label_set = label_set[100:199]
# 使用全新数据计算不同k值的误差值：(发现k=10的误差最小)
yHat01 = lwlr.lwlrTest(new_data_set, train_data_set, train_label_set,
0.1)
yHat1 = lwlr.lwlrTest(new_data_set, train_data_set, train_label_set, 1)
yHat10 = lwlr.lwlrTest(new_data_set, train_data_set, train_label_set,
10)
# 计算不同k值的误差值：
error01 = abalone.rssError(new_label_set, yHat01.T)
# error01 == 56.7987246777
print("\n\n\n error01 == %s" % (error01))
error1 = abalone.rssError(new_label_set, yHat1.T)
# error1 == 429.89056187
print("\n error1 == %s" % (error1))
error10 = abalone.rssError(new_label_set, yHat10.T)
# error10 == 549.118170883
print("\n error10 == %s" % (error10))
def test_lr(self):
dataMat, labelMat = lr.loadDataSet("ex0.txt")
w = lr.standRegres(dataMat, labelMat)
print("\n w == %s" % (w))
xMat = mat(dataMat)
yMat = mat(labelMat)
# 预测值
yHat = xMat * w
figure = plt.figure()
ax = figure.add_subplot(111)
ax.scatter(xMat[:, 1].flatten().A[0], yMat.T[:, 0].flatten().A[0])
xCopy = xMat.copy()
# 进行排序
xCopy.sort(0)
yHat1 = xCopy * w
print("\n xCopy[:, 1] == %s" % (xCopy[:, 1]))
x = xCopy[:, 1]
y = yHat1
# 画直线
ax.plot(x.A, y.A, "r")
plt.show()
def test_lwlr_plot(self):
dataMat, labelMat = lr.loadDataSet("ex0.txt")
# k = 1 则与线性回归基本一致（当k=0.01时，预测结果比较好）
lwlr.lwlrPlot(dataMat, labelMat, 0.01)