仅使用系数和截距模拟sklearn Logistic回归预报_proba

问题描述

我将创建虚拟数据并在其上训练sklearn Logistic回归。然后，我想获取predict_proba的输出，但仅使用自己的coef_和intercept_计算，但是结果不同。设置如下：

X = [[0,0],[0,1,2,[1,1],0]]
y = [0,2] 

# Fit the classifier
clf = linear_model.LogisticRegression(C=1e5,multi_class="ovr",class_weight="balanced")
clf.fit(X,y)

然后，我将仅使用有关Sigmoid和softmax的知识来获取输出：

softmax([
         expit(np.dot([[0,0]],clf.coef_[0]) + clf.intercept_[0]),expit(np.dot([[0,clf.coef_[1]) + clf.intercept_[1]),clf.coef_[2]) + clf.intercept_[2])
])

但是它将返回不同的值

clf.predict_proba([[0,0]])

array([[0.281399,0.15997556,0.55862544]])与array([[0.29882052],[0.24931448],[0.451865 ]])

相反

解决方法

from sklearn import linear_model
from scipy.special import expit,softmax
import numpy as np

# Data
X = [[0,0],[0,1,2,[1,1],0]]
y = [0,2]

# Classifier
clf = linear_model.LogisticRegression(C=1e5,multi_class="ovr",class_weight="balanced")
clf.fit(X,y)

# Predicted probabilities
print(clf.predict_proba([[0,0]]))
#[[0.281399 0.15997556 0.55862544]]

# Recalculated predicted probabilities without softmax
prob1 = np.array([expit(np.dot([[0,0]],clf.coef_[0]) + clf.intercept_[0]),expit(np.dot([[0,clf.coef_[1]) + clf.intercept_[1]),clf.coef_[2]) + clf.intercept_[2])]).reshape(1,-1)

print(prob1 / np.sum(prob1))
#[[0.281399 0.15997556 0.55862544]]

# Recalculated predicted probabilities with softmax
prob2 = np.log(prob1)

print(softmax(prob2))
#[[0.281399 0.15997556 0.55862544]]