Reviewing ensemble classification methods in breast cancer.

Area of Science:

• Bioinformatics and Computational Biology
• Medical Informatics
• Cancer Research

Background:

• Ensemble methods combine multiple techniques to enhance prediction accuracy, overcoming individual method limitations.
• These methods are increasingly applied in bioinformatics, particularly for complex diseases like breast cancer.
• Breast cancer research benefits from ensemble techniques due to its high prevalence and mortality rates.

Purpose of the Study:

• To systematically map the state-of-the-art in ensemble classification for breast cancer research.
• Analysis covers publication venues, medical tasks, research types, ensemble architectures, constituent techniques, validation, tools, and optimization.
• This systematic mapping study provides a comprehensive overview of current trends and identifies research gaps.

Main Methods:

• Systematic mapping study of 193 papers published from 2000 onwards.
• Data extracted from IEEE Xplore, ACM digital library, Scopus, and PubMed.
• Analysis focused on nine key aspects of ensemble classification in breast cancer research.

Main Results:

• Diagnosis is the most frequent medical task studied. Experiment-based empirical and evaluation-based research types dominate.
• Homogeneous ensemble types are most common. Decision trees, support vector machines, and artificial neural networks are frequently used base techniques.
• Wisconsin Breast Cancer dataset and k-fold cross-validation are prevalent for evaluation. Weka and R are common tools. Optimization of base techniques is infrequent.

Conclusions:

• The study identifies gaps and offers recommendations for breast cancer researchers.
• Most studies report improved accuracy with ensemble classifiers compared to single classifiers.
• Further systematic literature reviews and meta-analyses are recommended to confirm the superiority of ensemble methods.