Predictive Analysis of Polycystic Ovarian Syndrome using CatBoost Algorithm
Yash Rathod, Aryan Komare, Ruchita Ajgaonkar, Shruti Chindarkar, Gajanan Nagare, Neelam Punjabi, Yogesh Karpate
Abstract
The global prevalence of PCOS is estimated to be between 5.5% and 12.6% in women in the age group of 17–45 years. Depending on the diagnostic criteria used, the prevalence estimates in India range from 8.2% to 22.5%. PCOS has become an undeniable area of research today and experts are trying to find the most suited and accurate method of detection for this widely varied syndrome. To address this issue, we offer a method that can aid in the early diagnosis and prediction of PCOS treatment based on a set of optimal and minimal parameters using machine learning. We employed seven different machine learning classifiers, CatBoost, Random Forest, SVM, Logistic Regression, Bernoulli Naive Bayes, Decision Tree, and K-Nearest Neighbour, and compared their outcomes in terms of accuracy, precision, sensitivity, F1 score and AUC. We discovered CatBoost to have the highest and most trustworthy accuracy to predict whether a woman should seek medical help or not for PCOS.