Rebalance Probabilities
Start Timer
0:00:00
Suppose we are training a binary classification algorithm.
The outcome is imbalanced, with 99.8% of individuals in our sample having an outcome value of 0, and 0.2% having an outcome value of 1. To build a model, we down-sample the training data using a random sample of 1% of the individuals who have an outcome of 0 and keeping all individuals with an outcome of 1.
Now we train the model on the smaller sample and build a binary classifier that predicts the probability of an individual with a value of 1.
How would we then adjust our output probabilities to use this model on the total, imbalanced population?
.
.
.
.
.
.
.
.
.
Topic
Machine Learning
Tags
Classification Algorithms
Calibration Techniques and Processes
Category
Case Study
Position
Data Scientist
Difficulty
Medium
Success
N/A
Last asked
This question was asked on
Companies
What did you think?
Comments