Blood cancer prediction using leukemia microarray gene data and hybrid logistic vector trees model.

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Blood cancer diagnosis is complex, time-consuming, and expensive.
• Early detection is crucial for effective leukemia treatment.
• Current diagnostic methods require significant expert involvement and multiple tests.

Purpose of the Study:

• To develop an automated system for accurate blood cancer prediction.
• To enhance diagnostic efficiency and reduce costs.
• To improve upon existing machine learning models for leukemia diagnosis.

Main Methods:

• Utilized a leukemia microarray gene dataset with 22,283 genes.
• Applied ADASYN (Adaptive Synthetic Sampling) for dataset balancing.
• Employed Chi-squared (Chi2) for high-dimensional feature selection.
• Developed a hybrid classifier: Logistics Vector Trees (LVTrees), combining logistic regression, support vector classifier, and extra tree classifier.
• Performed k-fold cross-validation and T-tests for rigorous evaluation.

Main Results:

• The proposed LVTrees model, with ADASYN and Chi2, achieved a perfect 100% accuracy.
• Demonstrated superior performance compared to state-of-the-art methods.
• Statistical T-tests confirmed the significant efficacy of the approach.
• K-fold cross-validation validated the model's robustness and supremacy.

Conclusions:

• The developed automated system offers a highly accurate and efficient method for blood cancer prediction.
• The LVTrees classifier presents a significant advancement in machine learning applications for oncology.
• This approach has the potential to revolutionize early blood cancer diagnosis, making it more accessible and cost-effective.