Diet Recommendation for Poly Cystic Ovarian Syndrome of Indian Patients Using Multi-Attribute and Multi-Labeling Classifier

Abstract

Polycystic Ovarian Syndrome (PCOS) is a poorly understood, under-diagnosed, and under-treated condition, with prevalence ranging from 2.2 percent to 26 percent worldwide. A prevalent endocrine disorder in women of childbearing age, PCOS is a syndrome that results in the development of ovarian cysts and may ultimately cause infertility. Oily skin, darker acne scars, weight gain, hypertension, and irregular menstrual cycles are a few of the prevalent symptoms. According to these symptoms, we have taken the PCOS dataset from Kaggle which contains 541 records and 43 attributes including patient number and target class. This target class contains the labels as 1 for PCOS affected and 0 for normal women. This data was collected from 10 nearest hospital from Kerala in India. Among 43 features, we selected 21 essential features based on gynecologist suggestions and recommending a proper diet. This work focuses on the prevention of both PCOS patients and normal women by recommending diet through the construction of rule set. We transform the existing target label into a new multi-label target class that includes the dietary class by using this rule set. This system recommends a PCOS diet based on clinical data of the patients using machine learning techniques such as K-Nearest Neighbor, Decision Tree, Random Forest classifier, and Multi-Layer Perceptron. Based on a variety of evaluation indicators, the findings are analyzed, and the performance of the algorithms is validated. This type of analysis is useful for early prevention and safe recovery of PCOS patients by recommending nutrition diets.