Application of GRKM Clustering Algorithm for DNA Data Analysis Using Self‐Encoder Dimensionality Reduction

This study proposes an enhanced k ‐means clustering algorithm to analyze gene microarray data with high dimensionality and complexity. Our method employs a four‐layer autoencoder for feature vector extraction and introduces a new GRKM algorithm. The GRKM integrates three key components: Gray wolf optimization (GWO) for determining optimal cluster numbers ( K ‐value), random walk for selecting initial cluster centers, and Markov distance for measuring sample similarity. When tested on a public lung cancer DNA microarray dataset, the improved algorithm demonstrates significant improvements in evaluation metrics including silhouette coefficient, within‐cluster sum of squares, and Davies–Bouldin (DB) index compared to traditional methods. These enhancements result in better accuracy and interpretability, effectively supporting the identification of highly expressed and differentially expressed genes relevant to lung cancer treatment.