# -*- encoding: utf-8 -*- """ 8.8.1 估计器得分 """ # 分类估计器 from sklearn.datasets import load_iris from sklearn.neighbors import KNeighborsClassifier from sklearn.model_selection import train_test_split as tsplit from sklearn.metrics import accuracy_score X, y = load_iris(return_X_y=True) X_train, X_test, y_train, y_test = tsplit(X, y, test_size=0.1) knn = KNeighborsClassifier() knn.fit(X_train, y_train) y_pred = knn.predict(X_test) print(accuracy_score(y_test, y_pred)) # 使用准确性指标评价函数 print(knn.score(X_test, y_test)) # 直接使用测试集对训练效果做出准确性评价 # 回归估计器 from sklearn.svm import SVR from sklearn.model_selection import train_test_split as tsplit from sklearn import metrics X, y = load_iris(return_X_y=True) X_train, X_test, y_train, y_test = tsplit(X, y, test_size=0.1) svr = SVR() svr.fit(X_train, y_train)
def build_model(X, W, dropout=0):
h = X
for wi in W[:-1]:
h = tf.nn.dropout(tf.nn.relu(tf.matmul(h, wi)),1.0-dropout)
return tf.nn.relu(tf.matmul(h,W[-1]))
# initialization
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
shutil.rmtree('log',ignore_errors=True)
tf.reset_default_graph()
# load the data
trX,trY,trId = get_data('data/train.csv')
teX,teY,teId = get_data('data/test.csv',loady=False)
trX,vaX,trY,vaY,trId,vaId = tsplit(trX,trY,trId,test_size=0.3)
# define the parameters
size_in = trX.shape[1]
size_out = trY.shape[1]
size_h = [size_in,512,64,size_out]
batch_size = 128
# define tf symbolic variables
X = tf.placeholder("float", [None, size_in])
Y = tf.placeholder("float", [None, size_out])
W = [init_weights([hi,ho]) for hi,ho in zip(size_h[:-1],size_h[1:])]
# define the model & operations
Yt = build_model(X, W, dropout=0.3)
Yp = build_model(X, W, dropout=0.0)
movies["budgetReturn"] = movies["gross"] / movies["budget"]
#Creating a column 'isLong' which defines whether or not a movie is a long
#(greater than the standard 120 minutes)
movies["isLong"] = np.where((movies["runtime"] > 120), 1,0)
#Creating an analagous column 'isShort' which defines whether or not a movie is
# short (less than the standard 75 minutes minimum)
movies["isShort"] = np.where((movies["runtime"] < 75),1,0)
#Splitting Data into training and testing
movies_train, movies_test = tsplit(movies, test_size = 0.2, random_state = 1)
#Moving columns to more appropriate locations
movies = movies[['name','genre', 'rating' , 'country', 'company', 'isMajor',
'released', 'year', 'runtime', 'isLong', 'isShort',
'score', 'votes', 'star', 'director',
'budget', 'gross', 'budgetReturn']]
#Removing duplicates, none found with below function
movies.duplicated()
#Splitting data into features (indepedent variables) and response
#(dependet variable)
features1 = movies.loc[:,"name":"budget"]
features2 = movies.loc[:,"budgetReturn"]
features = pd.concat([features1, features2],axis = 1)
# -*- encoding: utf-8 -*-
"""
8.4.2 贝叶斯分类
"""
from sklearn.datasets import fetch_20newsgroups
from sklearn.feature_extraction.text import TfidfVectorizer # TF-IDF向量
from sklearn.naive_bayes import MultinomialNB # 导入多项式分布的朴素贝叶斯模型
from sklearn.model_selection import train_test_split as tsplit
from sklearn.metrics import classification_report # 导入分类结果评估报告函数
X, y = fetch_20newsgroups(return_X_y=True) # 获取新闻数据集和分类标签集
vectorizer = TfidfVectorizer()
vdata = vectorizer.fit_transform(X) # 文本转为TF-IDF向量
x_train, x_test, y_train, y_test = tsplit(vdata, y, test_size=0.1)
m = MultinomialNB() # 实例化多项式分布的朴素贝叶斯分类器
m.fit(x_train, y_train) # 模型训练
precision = m.score(x_test, y_test)
print('测试集分类准确率：%0.2f' % precision)
y_pred = m.predict(x_test)
report = classification_report(y_test, y_pred)
print('测试集分类结果报告：\n', report)