高级编程技术 sklearn课后习题

from sklearn import datasets
from sklearn import cross_validation
from sklearn.naive_bayes import GaussianNB
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier
from sklearn import metrics

# Datasets
dataset = datasets.make_classification(n_samples=1000, n_features=10)

# Inspect the data structures
print("dataset.data: \n", dataset[0])
print("dataset.target: \n", dataset[1])

# Cross-validation
kf = cross_validation.KFold(1000, n_folds=10, shuffle=True)
for train_index, test_index in kf:
    X_train, y_train = dataset[0][train_index], dataset[1][train_index]
    X_test, y_test = dataset[0][test_index], dataset[1][test_index]

# Inspect the data structures
print("X_train: \n", X_train)
print("y_train: \n", y_train)
print("X_test: \n", X_test)
print("y_test: \n", y_test)

# Navie Bayes
GaussianNB_clf = GaussianNB()
GaussianNB_clf.fit(X_train, y_train)
GaussianNB_pred = GaussianNB_clf.predict(X_test)

# Inspect the data structures
print("GaussianNB_pred: \n", GaussianNB_pred)
print("y_test: \n", y_test)

# SVM
SVC_clf = SVC(C=1e-01, kernel='rbf', gamma=0.1)
SVC_clf.fit(X_train, y_train)
SVC_pred = SVC_clf.predict(X_test)

# Inspect the data structures
print("SVC_pred: \n", SVC_pred)
print("y_test: \n", y_test)

# Random Forest
Random_Forest_clf = RandomForestClassifier(n_estimators=6)
Random_Forest_clf.fit(X_train, y_train)
Random_Forest_pred = Random_Forest_clf.predict(X_test)

# Inspect the data structures
print("Random_Forest_pred: \n", Random_Forest_pred)
print("y_test: \n", y_test)

# Performance evaluation
print()
GaussianNB_acc = metrics.accuracy_score(y_test, GaussianNB_pred)
print("  GaussianNB_acc: ", GaussianNB_acc)
GaussianNB_f1 = metrics.f1_score(y_test, GaussianNB_pred)
print("  GaussianNB_f1: ", GaussianNB_f1)
GaussianNB_auc = metrics.roc_auc_score(y_test, GaussianNB_pred)
print("  GaussianNB_auc: ", GaussianNB_auc)

print()
SVC_acc = metrics.accuracy_score(y_test, SVC_pred)
print("  SVC_acc: ", SVC_acc)
SVC_f1 = metrics.f1_score(y_test, SVC_pred)
print("  SVC_f1: ", SVC_f1)
SVC_auc = metrics.roc_auc_score(y_test, SVC_pred)
print("  SVC_auc: ",SVC_auc)

print()
Random_Forest_acc = metrics.accuracy_score(y_test, Random_Forest_pred)
print("  Random_Forest_acc: ", Random_Forest_acc)
Random_Forest_f1 = metrics.f1_score(y_test, Random_Forest_pred)
print("  Random_Forest_f1: ", Random_Forest_f1)
Random_Forest_auc = metrics.roc_auc_score(y_test, Random_Forest_pred)
print("  Random_Forest_auc: ", Random_Forest_auc)

运行结果：

步骤如下：

1、 生成数据集 

dataset = datasets.make_classification(n_samples=1000, n_features=10)；

2、 将数据集划分为训练集和测试集；

kf = cross_validation.KFold(1000, n_folds=10, shuffle=True)
for train_index, test_index in kf:
    X_train, y_train = dataset[0][train_index], dataset[1][train_index]
    X_test, y_test = dataset[0][test_index], dataset[1][test_index]

对于Naive Bayes（ 天真的港湾 朴素贝叶斯），SVM， Random Forest三种算法，接下来的套路都是相同的，不妨以朴素贝叶斯为例。

3、设定训练方式；

GaussianNB_clf = GaussianNB()

4、用训练集训练；

   GaussianNB_clf.fit(X_train, y_train)

5、生成预测集；

   GaussianNB_pred = GaussianNB_clf.predict(X_test)

6、比较预测集和测试集，给出正确性评估，有三种评估方式。

GaussianNB_acc = metrics.accuracy_score(y_test, GaussianNB_pred)
print("  GaussianNB_acc: ", GaussianNB_acc)
GaussianNB_f1 = metrics.f1_score(y_test, GaussianNB_pred)
print("  GaussianNB_f1: ", GaussianNB_f1)
GaussianNB_auc = metrics.roc_auc_score(y_test, GaussianNB_pred)
print("  GaussianNB_auc: ", GaussianNB_auc)

得分评估：

跑了几次程序，结果都不太一样，三种算法孰优孰劣很难讲，既然他们能被作为库函数，肯定有自己的独特优点。