Research on Bone Joint Cancer miRNA Recognition Algorithm Based on Isomap and Improved Bagging

Abstract

With the widespread adoption of intelligent diagnosis for genetic diseases in the medical field, miRNA sequence-based diagnostic methods have gradually become a focal point of contemporary biological research. Against this backdrop, this paper innovatively proposes a genomic recognition algorithm for bone joint cancer based on manifold dimensionality reduction and the improved Bagging framework. First, the high-dimensional miRNA sequences are assumed to lie in a low-dimensional manifold space, and feature extraction is performed using the Isomap algorithm. This step not only effectively resolves the "curse of dimensionality" caused by high-dimensional data but also successfully preserves key information for prediction. Second, considering that single models are prone to overfitting or underfitting, an ensemble learning framework is introduced for the prediction task. To prevent model redundancy, an improved Bagging algorithm based on mutual information selection is proposed. This algorithm dynamically adjusts the combination of sub-models, effectively reducing redundancy and improving prediction accuracy. Finally, empirical analysis on a bone joint cancer dataset demonstrates that the proposed method outperforms traditional baseline methods, showing greater competitiveness and superiority.