Synergistic feature selection and distributed classification framework for high-dimensional medical data analysis

Abstract

Feature selection and classification efficiency and accuracy are key to improving decision-making regarding medical data analysis. Since the medical datasets are large and complex, they give rise to certain problematic issues such as computational complexity, limited memory space, and a lesser number of correct classifications. In order to overcome these drawbacks, the new integrated algorithm is presented here: Synergistic Kruskal-RFE Selector and Distributed Multi-Kernel Classification Framework (SKR-DMKCF). The innovative architecture of SKR-DMKCF results in the reduction of dimensionality while preserving useful characteristics of the image utilizing recursive feature elimination and multi-kernel classification in a distributed environment. Detailed evaluations were performed on four broad medical datasets and established our performance advantage. The average feature reduction ratio was 89 % for the proposed method, SKR-DMKCF, which can outperform all the methods by achieving the best classification average accuracy of 85.3 %, precision of 81.5 %, and recall 84.7 %. On the efficiency calculations, it was seen that the memory usage is a 25 % reduction compared to the existing methods and the speed-up time was a significant improvement as well to assure scalability for resource-limited environments.

• •
  Innovative Synergistic Kruskal-RFE Selector for efficient feature selection in medical datasets.
• •
  Distributed Multi-Kernel Classification Framework achieving superior accuracy and computational efficiency.