Robust and stable feature selection by integrating ranking methods and wrapper technique in genetic data...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-dimensional data, such as microarray data, present computational challenges and reduce generalization accuracy in classification tasks.
• Microarray data is crucial for diagnosing diseases like cancer and tumors, but its high dimensionality necessitates dimension reduction.

Purpose of the Study:

• To develop a robust method for dimension reduction and classification of genetic microarray data.
• To enhance the stability and accuracy of feature selection and classification in high-dimensional genetic datasets.

Main Methods:

• A hybrid feature ranking method is proposed, fusing multiple techniques for robust feature selection.
• A wrapper method is integrated with the hybrid ranking to capture gene interactions.
• Support Vector Machine (SVM) is employed for classification, addressing class imbalance issues in training data.

Main Results:

• The proposed feature selection process demonstrated robustness within the interval of [0.70, 0.88].
• The classification accuracy achieved by the approach ranged from 91% to 96% across five microarray databases.

Conclusions:

• The developed method effectively addresses dimension reduction and classification challenges in microarray data.
• The findings highlight the potential of the proposed approach for accurate disease diagnosis using genetic data.