Scikit-Learn 機器學習筆記 – SVM

---

參考文檔： handson-ml

---

import numpy as np# 加載鳶尾花數(shù)據(jù)集 def load_dataset():from sklearn import datasetsiris = datasets.load_iris()# print(iris)# 使用第3和第4個特征X = iris['data'][:, (2, 3)]# bool類型轉(zhuǎn)為數(shù)值型y = (iris['target'] == 2).astype(np.float64)return X, y# 線性SVM二分類器 def linear_svm_classify(X, y):from sklearn.preprocessing import StandardScalerfrom sklearn.svm import LinearSVC# 特征縮放scale = StandardScaler()X_scaled = scale.fit_transform(X.astype(np.float64))# C是損失函數(shù)項的系數(shù)，稱為懲罰系數(shù)svm_clf = LinearSVC(C=1, loss='hinge')svm_clf.fit(X_scaled, y)# 預測predict = svm_clf.predict([[5.5, 1.7]])print('線性SVM二分類器預測為：', predict)# 非線性SVM二分類器、 def nonlinear_svm_classify(X, y):from sklearn.preprocessing import PolynomialFeatures, StandardScalerfrom sklearn.svm import LinearSVC# 增加多項式特征poly_features = PolynomialFeatures(degree=3)X_poly = poly_features.fit_transform(X.astype(np.float64))# 特征縮放scale = StandardScaler()X_scaled = scale.fit_transform(X_poly)# 需要指定損失函數(shù)poly_svm_clf = LinearSVC(C=10, loss='hinge')poly_svm_clf.fit(X_scaled, y)# 預測，先把樣本特征向量轉(zhuǎn)為包含多項式特征得向量sample_poly = poly_features.fit_transform([[5.5, 1.7]])predict = poly_svm_clf.predict(sample_poly)print('非線性SVM二分類器預測為：', predict)# 多項式核SVM二分類器 def kernel_poly_svm_classify(X, y):from sklearn.preprocessing import StandardScalerfrom sklearn.svm import SVC# 特征縮放scale = StandardScaler()X_scaled = scale.fit_transform(X)# 使用多項式核，kernel_poly_svm_clf = SVC(kernel="poly", degree=3, coef0=1, C=5)kernel_poly_svm_clf.fit(X_scaled, y)# 預測，特征向量不用轉(zhuǎn)換predict = kernel_poly_svm_clf.predict([[5.5, 1.7]])print('核SVM二分類器預測為：', predict)# 高斯RBF核SVM二分類器 def kernel_rbf_svm_classify(X, y):from sklearn.preprocessing import StandardScalerfrom sklearn.svm import SVC# 特征縮放scale = StandardScaler()X_scaled = scale.fit_transform(X)# 增大gamma和減小C都會加大擬合kernel_rbf_svm_clf = SVC(kernel="rbf", gamma=5, C=1)kernel_rbf_svm_clf.fit(X_scaled, y)# 預測，特征向量不用縮放predict = kernel_rbf_svm_clf.predict([[5.5, 1.7]])print('高斯RBF核SVM二分類器預測為：', predict)if __name__ == '__main__':# 加載數(shù)據(jù)集X, y = load_dataset()# 線性SVM二分類器# linear_svm_classify(X, y)# 非線性SVM二分類器# nonlinear_svm_classify(X, y)# 多項式核SVM二分類器kernel_poly_svm_classify(X, y)# 高斯rbf核SVM二分類器kernel_rbf_svm_classify(X, y)

---

與50位技術(shù)專家面對面20年技術(shù)見證，附贈技術(shù)全景圖

總結(jié)

以上是生活随笔為你收集整理的Scikit-Learn 机器学习笔记 -- SVM的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。