Purpose: Polycystic ovary syndrome (PCOS) affects a noteworthy number from the age of fertile women, presenting as a common hormonal disorder. It leads to irregular hormone levels and can develop as an ovarian cyst. Methods: Diagnosing PCOS can be difficult in real-world situations because it is heavily dependent on the skills of doctors. In this study, an innovative approach suggested the proposal named SVEMI (Soft Voting Ensemble with Mutual Information) for feature selection and classification through machine learning, utilizing four ML models as a base learner. SVEMI combines the strengths of soft voting ensemble techniques and mutual information-based feature selection to achieve high predictive performance with a reduced feature set using three categories of feature selection procedures. Results:The experimental results, derived from 20 selected features, shows that SVEMI method achieved an impressive accuracy of 96.79% on a real-world dataset, surpassing previous methods. Conclusions: Hence, the results obtained from SVEMI presents a promising strategy for early detection of PCOS.