A weighted ensemble-based active learning model to label microarray data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Classifying cancerous genes from microarray data is crucial but labeling is costly.
• Existing methods require substantial labeled data, limiting scalability.

Purpose of the Study:

• To develop a semi-supervised active learning model for efficient microarray data labeling.
• To improve the accuracy of cancerous gene classification with reduced labeling effort.

Main Methods:

• An ensemble approach combining three classifiers for consensus prediction.
• Active learning strategies to select informative instances from unlabeled data.
• Optimizing the search space within a sparse learning pool.

Main Results:

• The proposed model achieves performance comparable to state-of-the-art methods.
• Demonstrated effectiveness on 10 diverse microarray datasets.
• Code and datasets are publicly available for reproducibility.

Conclusions:

• The ensemble-based active learning framework offers an efficient solution for microarray data classification.
• This approach significantly reduces the cost and effort associated with data labeling.
• Enables accurate predictions even with limited labeled genomic data.