Related Experiment Videos
Generating concise and accurate classification rules for breast cancer diagnosis.
1School of Computing, National University of Singapore, Kent Ridge, Singapore, Singapore. rudys@comp.nus.edu.sg
Artificial Intelligence in Medicine
|February 17, 2000
Summary
This study enhances neural network accuracy for breast cancer diagnosis by refining data preprocessing. Improved data selection and cleaning yield more accurate and concise classification rules.
Area of Science:
- Computer Science
- Medical Informatics
- Artificial Intelligence
Background:
- Previous work introduced an algorithm for extracting classification rules from neural networks.
- The algorithm was previously applied to breast cancer diagnosis.
Purpose of the Study:
- To improve the accuracy of neural networks and their extracted classification rules.
- To enhance the rule extraction process through data preprocessing.
Main Methods:
- Implementing a data preprocessing step before rule extraction.
- Selecting relevant input attributes.
- Removing samples with missing attribute values.
Main Results:
- The preprocessing method improved the accuracy of both the neural networks and the extracted rules.
- Generated rules were more concise and accurate compared to existing methods.
- Enhanced performance in breast cancer diagnosis applications.
Conclusions:
- Simple data preprocessing significantly boosts neural network and rule extraction accuracy.
- The refined method offers a more effective approach for medical diagnosis and rule-based systems.
- This technique provides more concise and accurate classification rules than other reported methods.