有约束条件的加权最小二乘的实现加权最小二乘原理Python实现

137 阅读 0 评论 91 点赞

我是靠谱客的博主笨笨啤酒，这篇文章主要介绍有约束条件的加权最小二乘的实现加权最小二乘原理Python实现，现在分享给大家，希望可以做个参考。

@[TOC] 有约束条件的加权最小二乘的实现

加权最小二乘原理

详细见网址：加权最小二乘法

Python实现

复制代码

import numpy as np
import pandas as pd
from scipy.optimize import minimize
class Constrained_regression:
def __init__(self, weight = None, intercept = True):
"""
默认选择最小二乘
"""
self.weight = weight
self.intercept = intercept
def weight_error(self, B, X_train, y_train):
# 加权最小二乘的损失函数
# 也是目标函数
# B表示beta，为待优化参数
data1 = np.dot(X_train, B)
data2 = y_train - data1
if self.weight is None: # 此时为普通最小二乘
self.weight = np.identity(self.sample_num)
weight_1 = np.linalg.pinv(self.weight)
temp3 = data2.transpose()
error1 = np.dot(temp3, weight_1)
error = np.dot(error1, data2)
return error
def cons(self):
# 创建约束条件
if self.intercept == True:
cons = ({'type': 'eq', 'fun': lambda x: np.sum(x[:-1])- 1})
else:
cons = ({'type': 'eq', 'fun': lambda x: np.sum(x)- 1})
bnds = []
for i in range(self.feature_num):
bnds.append((0, 1))
if self.intercept == True:
bnds.append((None, None))
bnds = tuple(bnds)
return cons, bnds
def fit(self, X, y):
# 模型训练
X_train = X
y_train = y
sample_num = X_train.shape[0]
feature_num = X_train.shape[1]
self.feature_num = feature_num
self.sample_num = sample_num
# 创造截距项
if self.intercept == True:
X_train['inter'] = 1
B_ini = list(np.zeros(feature_num) + 1/feature_num)
B_ini.append(0.5)
fun = lambda x: self.weight_error(B=x, X_train=X_train, y_train=y_train)
cons, bnds = self.cons()
res = minimize(fun, B_ini, method='SLSQP', constraints=cons, tol=1e-5, bounds = bnds)
return res.x
if __name__ == "__main__":
data = pd.read_csv('YFDdata.csv')
X_train = data[['bond', 'Conbond', 'Credebt']]
y_train = data['YFD']
model = Constrained_regression()
conf = model.fit(X_train, y_train)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
import numpy as np
import pandas as pd
from scipy.optimize import minimize
class Constrained_regression:
def __init__(self, weight = None, intercept = True):
"""
默认选择最小二乘
"""
self.weight = weight
self.intercept = intercept
def weight_error(self, B, X_train, y_train):
# 加权最小二乘的损失函数
# 也是目标函数
# B表示beta，为待优化参数
data1 = np.dot(X_train, B)
data2 = y_train - data1
if self.weight is None: # 此时为普通最小二乘
self.weight = np.identity(self.sample_num)
weight_1 = np.linalg.pinv(self.weight)
temp3 = data2.transpose()
error1 = np.dot(temp3, weight_1)
error = np.dot(error1, data2)
return error
def cons(self):
# 创建约束条件
if self.intercept == True:
cons = ({'type': 'eq', 'fun': lambda x: np.sum(x[:-1])- 1})
else:
cons = ({'type': 'eq', 'fun': lambda x: np.sum(x)- 1})
bnds = []
for i in range(self.feature_num):
bnds.append((0, 1))
if self.intercept == True:
bnds.append((None, None))
bnds = tuple(bnds)
return cons, bnds
def fit(self, X, y):
# 模型训练
X_train = X
y_train = y
sample_num = X_train.shape[0]
feature_num = X_train.shape[1]
self.feature_num = feature_num
self.sample_num = sample_num
# 创造截距项
if self.intercept == True:
X_train['inter'] = 1
B_ini = list(np.zeros(feature_num) + 1/feature_num)
B_ini.append(0.5)
fun = lambda x: self.weight_error(B=x, X_train=X_train, y_train=y_train)
cons, bnds = self.cons()
res = minimize(fun, B_ini, method='SLSQP', constraints=cons, tol=1e-5, bounds = bnds)
return res.x
if __name__ == "__main__":
data = pd.read_csv('YFDdata.csv')
X_train = data[['bond', 'Conbond', 'Credebt']]
y_train = data['YFD']
model = Constrained_regression()
conf = model.fit(X_train, y_train)