保存模型,加载模型
from sklearn.externals import joblib
1.保存jobib.dump(rf,‘test.plk’)
sgd = SGDRegressor()
sgd.fit(x_train, y_train)
print(sgd.coef_)
# 保存训练好的模型
joblib.dump(sgd, './tmp/test.pkl')
2.加载estimator = joblib.load(‘test.plk’)
modle = joblib.load('./tmp/test.pkl')
y_predict = std_y.inverse_transform(modle.predict(x_test))
print('保存模型预测价格:',y_predict)
逻辑回归-----解决2分类的利器
解决2分类问题:
线性回归的式子作为输入:通过sigmoid函数
均方误差:不存在多个局部最低点,只有一个最小值,
对数似然损失:多个局部最小值
净量改变:1.多次随机初始化,多次比较最小值结果
2.求解过程中,调整学习率
- 逻辑回归,不需要做目标值标准化的原因是,分类问题不是回归问题,回归问题要做标准化的原因是?预测的结果是标准化之后的结果,目标值不做标准化,相差很大
1.哪个类别少,判定概率值就选这个
import pandas as pd
import numpy as np
from sklearn.metrics import classification_report
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
def logistic():
"""
retrun:逻辑回归做二分类问题预测
"""
column = ['Sample code number', 'Clump Thickness', 'Uniformity of Cell Size','Uniformity of Cell Shape','Marginal Adhesion', 'Single Epithelial Cell Size','Bare Nuclei','Bland Chromatin','Normal Nucleoli','Mitoses','Class']
data = pd.read_csv('./bate/21.csv', names=column)
# print(data)
# 缺失值进行处理
data = data.replace(to_replace='?', value=np.nan) # 用到了np函数的替换
# 缺失值用,把控制删除
data = data.dropna()
# 进行数据分割
x_train, x_test, y_train, y_test = train_test_split(data[column[1:10]], data[column[10]], test_size=0.25) # from sklearn.model_selection import train_test_split
# 进行标准化
std = StandardScaler() # from sklearn.preprocessing import StandardScaler
x_train = std.fit_transform(x_train)
x_test = std.transform(x_test)
# 逻辑回归预测
lg = LogisticRegression(C=1.0)# from sklearn.linear_model import LogisticRegression
lg.fit(x_train, y_train) # 不断优化w值
print(lg.coef_) # 权重参数,9个特征的系数
print("准确率:",lg.score(x_test, y_test))
y_predict = lg.predict(x_test)
print('&'*50)
print('召回率:',classification_report(y_test, y_predict, labels=[2, 4], target_names=["良性", "恶性"])) # 目标值 2代表良性
return None
if __name__== '__main__':
logistic()