Multistage feature selection and stacked generalization model for cancer detection.

Area of Science:

• Biomedical Engineering
• Computational Biology
• Medical Informatics

Background:

• Reliable cancer screening is crucial for early detection and improved patient outcomes.
• Current methods may require extensive feature sets, impacting efficiency and accuracy.
• The need for robust and efficient diagnostic tools in oncology is paramount.

Purpose of the Study:

• To develop a novel approach for cancer screening by integrating intelligent feature selection with a stacking classifier.
• To reduce the number of features required for cancer diagnosis without compromising accuracy.
• To enhance the performance metrics of cancer identification models.

Main Methods:

• A stacking classifier was developed, combining Logistic Regression, Naïve Bayes, Decision Tree, and a Multilayer Perceptron as a meta-classifier.
• An intelligent feature selection technique was employed to identify the most critical features for cancer diagnosis.
• The proposed method was evaluated on benchmark datasets to assess its diagnostic performance.

Main Results:

• The proposed method significantly reduced the number of features needed for cancer screening.
• The stacking model achieved superior performance across key metrics including accuracy, sensitivity, precision, specificity, and Area Under the Curve (AUC).
• An outstanding 100% accuracy, sensitivity, specificity, and AUC were attained using the selected optimal feature subsets.

Conclusions:

• Intelligent feature selection is a key factor in improving the performance of cancer identification models.
• The proposed stacking classifier with optimized feature selection offers a highly accurate and efficient solution for cancer screening.
• This approach simplifies the process of cancer identification, making it more accessible and reliable.