机器学习笔记

# 标准化处理
from sklearn.preprocessing import StandardScaler
X_norn = StandardScaler().fit_transform(X)

# 数据降维, PCA 
from sklearn.decomposition import PCA
pca = PCA(n_components=要降维到几个维度)
X_pca = pca.fit_transform(标准化后的数据X)
# 获取各维度的比值
var_ratio = pca.explained_variance_ratio_

# 线性回归
from sklearn.linear_model import LinearRegression
lr1 = LinearRegression()
lr1.fit(X_train, y_train)

# 数据生成
	# 生成若干个一维数据点
	np.linspace(开始值, 结束值, 生成几个点).reshape(-1, 1)
	# 生成二维数据
	xx, yy = np.meshgrid(np.arange(开始值, 结束值, 间隔一定距离生成点), np.arange(开始值, 结束值, 间隔一定距离生成点))
	# 生成若干行, 两列
	x_range = np.c_[xx.ravel(), yy.ravel()]


# R2
# 越接近1越好
from sklearn.metrics import r2_score
r2 = r2_score(y, y_predict)


# 把单项式数据转化为多项式数据
# fit_transform只要调用一次, 之后的转换, 用transform就可以
from sklearn.preprocessing import PolynomialFeatures
poly = PolynomialFeatures(degree=最高几次)
X_2_train = poly.fit_transform(X_train)
X_2_test = poly.transform(X_test)


# 检测异常点
from sklearn.covariance import EllipticEnvelope
# 0.02越大, 检测到的异常点越多
model = EllipticEnvelope(contamination=0.02)
model.fit(X)
# y_predict_bad是-1的, 表示异常点
y_predict_bad = model.predict(X)


# 数据分离
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=4, test_size=0.4)


# 混淆矩阵
from sklearn.metrics import confusion_matrix
cm = confusion_matrix(y_test, y_test_prodict)
TP = cm[1, 1]
TN = cm[0, 0]
FP = cm[0, 1]
FN = cm[1, 0]
# 准确率: 整体样本中, 预测正确样本数的比例
Accuracy = (TP + TN)/(TP + TN + FP + FN)
# 灵敏度(召回率): 正样本中, 预测正确的比例
Sensitivity = Recall = TP/(TP + FN)
# 特异度: 负样本中, 预测正确的比例
Specificity = TN/(TN + FP)
# 精确率: 预测结果为正的样本中, 预测正确的比例
Precision = TP/(TP + FP)
# F1分数: 综合Precision和Recall的一个判断指标
F1 Score = 2*Precision * Recall/(Precision + Recall)