Logistic Regression with continuous data using sklearn in python

Question

How can I describe these points with a regression? In the example the LinearRegression doesn't fit the logistic distribution of the points. The LogisticRegression() from sklearn just accept binary data. My y-values are continuous from 0 to 1. Do I have to transform the data or how do I get a appropriate model?

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from sklearn.linear_model import LinearRegression
from sklearn.linear_model import LogisticRegression

a = np.array([1,2,3,4,5,6,7,8,9,10,11,12,13,14])
b = [0,0,0.01,0.08,0.16,00.28,0.5,0.66,0.8,0.9,0.95,0.99,1,1]
data = pd.DataFrame({'x': a, 'y':b})

LM = LinearRegression()
LM.fit(data[["x"]],data[["y"]])

plt.scatter(a,b)
plt.plot([1,14], LM.predict([[1],[14]]), color = "red")
plt.show() 

LogM = LogisticRegression()
LogM.fit(data[["x"]],data[["y"]])  # doesn't work

scatter plot with linear model

Tristan Nemoz · Accepted Answer · 2020-06-27 16:36:23Z

A logistic regression is generally used to classify labels, even though it outputs a real between 0 and 1. This is why sklearn wants binary data in y: so that it can train the model.

In your case, you have a sigmoid function s(x)=1/(1+exp(alpha*x + beta)) and you want to find alpha and beta. I think the simplest way to do this is first of all to transform your data:

new_a = a[2:-2]
new_b = np.array(b[2:-2]) # Getting rid of 0 and 1 values
new_b = np.log((1 / new_b) - 1)

Now new_b is an array whose values are under the form alpha*new_a + beta, so you can train a LinearRegression model on it to find alpha:

model = LinearRegression()
model.fit(new_a.reshape(-1, 1), new_b.reshape(-1, 1))
alpha = model.coef_[0, 0]
beta = l.predict([[0]])[0, 0]

Finally, you can see test whether this correesponds to what you expect:

predicted = 1 / (1 + np.exp(alpha * a + beta))
plt.figure()
plt.plot(a, b)
plt.plot(a, predicted)
plt.show()

Collectives™ on Stack Overflow

Logistic Regression with continuous data using sklearn in python

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related