回归预测之SVR——运行+调优

需要注意的是“数据归一化处理”+“核函数的选择”

from sklearn.svm import SVR

# 特征数据归一化处理
from sklearn.preprocessing import StandardScaler

stdScaler = StandardScaler()
X_train = stdScaler.fit_transform(X_train) # 训练集数据
X_test = stdScaler.transform(X_test) #区别于上一行

svr_reg = SVR(
    kernel='linear'
)

svr_reg.fit(X_train,y_train)

y_pred=svr_reg.predict(X_test)
print(y_pred)

[ 49.70912563 4.8657277 11.89035597 19.68675085 23.49184558
-29.85398429 -3.21393278 81.32381918 -85.61315983 66.01787202
-69.82674359 -25.38715002 -40.24706408 -50.64749623 11.60096158
-26.06522083 -10.14868146 -16.73690997 -18.80629055 -37.52370137
6.86249293 94.97742697 14.0344571 37.52105985 61.02493142
-7.51511012 42.18587884 89.75971582 47.44765728 14.08536618]

plt.figure(figsize=(15, 10))

# 创建t变量
t = np.arange(len(X_test))
# 绘制y_test曲线
plt.plot(t, y_test, 'r', linewidth=2, label='真实值')
# 绘制y_pred曲线
plt.plot(t, y_pred, 'g', linewidth=2, label='预测值')
# 设置图例
plt.legend()
plt.show()

svr_reg.score(X_test,y_test)

metrics.mean_squared_error(y_test, y_pred)

0.9999890573808796

0.022024646328507425