Predicting Breast Cancer Leveraging Supervised Machine Learning Techniques.

Area of Science:

• Oncology
• Computer Science
• Biomedical Engineering

Background:

• Breast cancer is a leading cause of death in women globally.
• Accurate diagnosis is challenging due to complex cell structures (microcalcifications, masses).
• Existing computer-aided diagnosis (CAD) systems require improvement for reliable breast cancer detection.

Purpose of the Study:

• To develop and validate a novel machine learning framework for precise breast cancer prediction.
• To enhance early and accurate detection of breast cancer using advanced computational methods.
• To improve upon existing diagnostic systems by leveraging machine learning.

Main Methods:

• Utilized machine learning algorithms: Random Forest, Gradient Boosting, Support Vector Machine, Artificial Neural Network, and Multilayer Perception.
• Employed a hybrid Multilayer Perceptron Model (MLP) with 5-fold cross-validation on the Wisconsin Diagnostic Breast Cancer (WDBC) dataset.
• Implemented a connection-based feature selection technique to eliminate recursive features for improved classification.

Main Results:

• Achieved a high classification accuracy of 99.12% on the WDBC dataset.
• Validated the framework's effectiveness on the Wisconsin Prognostic (WPBC) and Wisconsin Original Breast Cancer (WOBC) datasets.
• Demonstrated the impact of efficient data preprocessing and feature selection on diagnostic accuracy.

Conclusions:

• The proposed machine learning framework significantly improves breast cancer prediction accuracy.
• The hybrid MLP model with advanced feature selection offers a robust solution for early breast cancer detection.
• This approach holds potential for enhancing radiologists' diagnostic capabilities and reducing diagnostic errors.