Polycystic ovary syndrome (PCOS) is a prevalent endocrine and metabolic disorder affecting women of reproductive age. Oxidative stress (OS) is suggested to play a significant role in the development of PCOS. Using antioxidants to reduce OS and maintain a healthy balance in the body could be a novel treatment approach for PCOS. This study analyzed transcriptome data from the Gene Expression Omnibus database, focusing on genes associated with OS. By implementing two machine learning algorithms, three OS-related biomarkers-HMOX1, MMP9, and KLF2-were successfully identified. To evaluate the diagnostic potential of these biomarkers, a Logistic regression model was employed. Additionally, granulosa cells were collected from healthy individuals and infertile women with PCOS, and the reliability of HMOX1, MMP9, and KLF2 was verified by quantitative real-time PCR experiments. Furthermore, small molecule drugs targeting proteins encoded by genes HMOX1 and MMP9 were predicted through the Drug Signature Database. Molecular docking of drugs to proteins identified two antioxidants, butein and demethoxycurcumin, as potential candidates for PCOS therapy.