Answers are grounded in this paper and cite their sources.
Cancer subtype analysis faces challenges due to limited availability of gene samples and the complexity of cancer gene expression data. The imbalance of Positive and negative category ratio and high-dimensional redundant information degrade prediction performance. This paper proposes an integrated extreme random forest with feature selection model TreeEM(Tree-enhanced Ensemble Model combining with feature selection) to enhance prediction ability and reduce computational costs. The TreeEM model combines the Max-Relevance and Min-Redundancy(MRMR) feature selection method with improved fusion undersampling random forest and extreme tree forest. The TreeEM model achieves excellent performance on three cancer datasets, especially on the multi-omics datasets BRCA(Breast Cancer) and ARCENE datasets, with average improvements of 7.90% and 1.90% in prediction accuracy, respectively. This model also uses TCGA data with known survival time for survival analysis and prediction, demonstrating the reliability of the TreeEM model. This work contributes to advancements in computational tools for cancer research, facilitating precision medicine approaches and improving decision-making. The above results provide new ideas for cancer subtype classification, but the existing methods still have limitations in data imbalance and high-dimensional feature processing. In the following section, the shortcomings of the current research and the innovative solutions of this paper are systematically described.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: