Polycystic Ovary Syndrome (PCOS) is an endocrine disorder affecting women in their reproductive years. This study is aimed at developing a performant and transparent machine learning model to assist doctors and health care workers in diagnosing PCOS. To this end, we apply several data split ratios, feature selection techniques and machine learning algorithms on a dataset containing 541 samples and 43 features. The best performing framework was the Mutual Information-based feature selection paired with an 85:15 data split ratio. Random Forest trained on this framework achieved an accuracy, recall, precision, AUC-ROC and f-1 score of 96, 96, 97, 98 and 96%, and is the best performing model. Several Explainable AI techniques are applied to the best performing RF model to make its decision-making more interpretable and transparent. These methods identify the most significant local and global features responsible for classifier predictions. While PCOS is a prevalent health issue around the world, there is a lack of knowledge about PCOS among women in India. In this paper, we develop PCOS_HelpBot, a question–answering chatbot prototyped using the open–source Llama2 large language model to address the need for improved education and awareness about PCOS. Such an application can assist doctors with diagnostic support, patient education and engagement.