Addressing Class Imbalance in Bayesian Classification Through Posterior Probability Adjustment.

Area of Science:

• Machine Learning
• Statistical Modeling
• Data Science

Background:

• Class imbalance is a common problem in classification tasks.
• Imbalanced datasets can cause predictive bias towards majority classes.
• Existing methods may not adequately address this bias.

Purpose of the Study:

• To introduce a novel Bayesian framework to mitigate bias from imbalanced datasets.
• To adjust posterior probabilities for more accurate classification.
• To propose a method that scales probabilities based on data representation.

Main Methods:

• Developed a straightforward Bayesian framework.
• Proposed a novel probability scaling technique.
• Adjusted posterior probabilities based on training data proportions.

Main Results:

• The proposed method effectively counteracts bias caused by imbalanced data.
• Posterior probabilities are scaled according to class representation.
• Achieved more balanced predictions in imbalanced classification tasks.

Conclusions:

• The novel Bayesian framework offers a robust solution for class imbalance.
• Scaling posterior probabilities based on data representation is effective.
• This approach improves classification accuracy on imbalanced datasets.