A hybrid machine learning feature selection model-HMLFSM to enhance gene classification applied to multiple colon...

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Colon cancer is a major global health concern requiring early detection for improved outcomes.
• Traditional colon cancer screening methods like colonoscopies are invasive.
• Machine learning offers a non-invasive approach for colon cancer classification using genetic data.

Purpose of the Study:

• To develop an improved machine learning model for non-invasive colon cancer gene classification.
• To address limitations of traditional machine learning models in handling high-dimensional genetic data and variable gene expression.
• To enhance the accuracy of colon cancer detection through effective feature selection.

Main Methods:

• Proposed a hybrid feature selection model (HMLFSM) for colon cancer gene classification.
• Implemented a two-phase feature selection approach combining Information Gain (IG) with Genetic Algorithms (GA), and minimum Redundancy Maximum Relevance (mRMR) with Particle Swarm Optimization (PSO).
• Tested the model on three distinct colon cancer genetic datasets.

Main Results:

• The HMLFSM model achieved significant accuracy improvements on colon cancer datasets, reaching ~95%, ~97%, and ~94%.
• The model effectively identified important and relevant genes while eliminating irrelevant ones.
• Demonstrated that selective input feature extraction is crucial for enhancing predictive performance in colon cancer gene analysis.

Conclusions:

• The proposed HMLFSM model significantly enhances colon cancer gene classification accuracy.
• The hybrid feature selection strategy effectively addresses challenges in high-dimensional genetic data.
• This approach holds promise for more accurate and non-invasive colon cancer detection.