The identification of genetic markers such as single nucleotide polymorphisms (SNPs) plays a critical role in understanding disease susceptibility and guiding personalized medicine. Recent advances in machine learning (ML) have provided new methods to address the complexities of genetic data. This paper introduces a novel ensemble learning technique that integrates attention-based neural networks with traditional random forest algorithms to enhance the identification of SNPs linked to disease outcomes. Using a benchmark dataset of genome-wide association studies (GWAS), we demonstrate how the proposed method improves prediction accuracy and model interpretability, thereby offering potential applications in clinical genomics.