Machine learning for diagnosis of polycystic ovarian syndrome (PCOS/PCOD)

A medical illness called polycystic ovary syndrome (PCOS) affects hormones and affects women who are in their childbearing age. A monthly cycle that is delayed or even missing is the result of hormonal imbalance. There is no recognized cause for PCOS. The most common symptoms of PCOS in women include excessive weight gain, facial hair development, acne, hair loss, skin discoloration, and irregular cycles, all of which can make it difficult to conceive. The likelihood of developing long-term issues like diabetes type 2 and heart problems may be reduced with weight management, early detection, and treatment. For early-stage detection and prediction, the current techniques and treatments fall short. To solve this issue, we provide a method that can predict the effectiveness of PCOS therapy based on an ideal and minimal set of criteria. We are using different machine learning algorithms and comparing the results of every algorithm to provide an efficient method for early detection of PCOS with the dataset collected from Kaggle. This paper, has two different datasets used for classification. The phase one involves finding the most important features causing PCOS and predicting the disorder using different machine learning algorithms using SMOTE based approach and without SMOTE approach. The phase two involves using the image dataset and classifying the images to be normal or PCOS using CNN.