From the survey it has been seen that upto 18% of woman all around the world of reproductive age suffer by a well-known reproductive endocrinopathies disease known as polycystic ovarian syndrome (PCOS). The earliest possible diagnosis and treatment of this condition has drawn research interest. Considering the serious imbalance of PCOS detection datasets will result in low classification performance and difficulty to detect the disease accurately and efficiently. In this study, the performance of two oversampling methods (SMOTE and ADASYN) is examined in conjunction with the RF classifier model in order to considerably increase the model's performance and evaluation metrics. The suggested PCOS prediction model's framework is made up of three distinct levels. The most important features are chosen in the first layer utilizing correlation and the principle component analysis (PCA) technique. The suggested model is trained in the second layer, and in the third layer, its performance is assessed in terms of classification accuracy (CA), precision, recall (sensitivity), Matthews' correlation coefficient (MCC), and area under the ROC curve (AUROC).The proposed RF+ADASYN algorithm clearly outperforms its counterparts and achieves a remarkable accuracy of CA, F1 Score, Sensitivity, AUROC, MCC, Training time and Prediction time are 97.94, 93.02, 94.19, 92.89, 0.90825, 0.7714, 0.139 sec and 0.005 sec respectively. The acquired simulation results demonstrate the excellence and efficacy of our suggested model.