Weighted Bayesian Belief Network: A Computational Intelligence Approach for Predictive Modeling in Clinical Datasets.

Area of Science:

• Computational biology
• Medical informatics
• Oncology

Background:

• Delayed breast cancer detection and treatment contribute significantly to mortality.
• Effective computational models are crucial for timely patient and physician intervention.
• Existing predictive models require enhancement for improved accuracy and reliability.

Purpose of the Study:

• To develop and evaluate a Weighted Bayesian Belief Network (WBBN) for breast cancer prediction.
• To utilize an automated ranking method for attribute weighting based on disease impact.
• To establish strong association rules for robust model construction.

Main Methods:

• Utilized the UCI breast cancer dataset for model training and testing.
• Employed an automated ranking method for attribute value pair weighting.
• Applied weighted association rule mining to identify attribute relationships and generate rules.
• Constructed the WBBN model using the Open Markov tool for structure and parametric learning.
• Trained the model on 70% of data and tested on 30%, with support ≥36% and confidence ≥70%.

Main Results:

• The WBBN model achieved a high accuracy of 97.18%.
• Weighted Bayesian confidence and lift measures were used to generate strong predictive rules.
• The WBBN model demonstrated superior performance compared to other predictive models in most evaluated cases.

Conclusions:

• The developed WBBN model offers a highly accurate computational approach for breast cancer prediction.
• This method can significantly aid in early detection and timely management of breast cancer.
• The study contributes to advancing breast cancer research and improving treatment quality.