Feature Selection And Logistic Regression For Optimal Detection Of Pcos

Authors

  • Valli Madhavi Koti, Y Venkateswarlu

Abstract

Premature abortion, infertility, anovulation are a major problem in today's world. Polycystic ovarian syndrome (PCOS) is found to have an important effect on the cause of infertility in reproductive age of women. Over five million women worldwide are having PCOS.  This disease causes ovarian dysfunction amplifying risks of abortion and infertility. Obesity, menstrual cycle irregularities, and overproduction of the male hormone, obesity, and hirsutism are the signs of PCOS. The variability of signs and the presence of various related gynaecological disorders make it challenging to identify PCOS. To acknowledge this issue, this study offers optimum and promisingly clinical and metabolic parameters for timely detection and prediction of PCOS at an early stage. Data from the patient’s survey of 541 women during physician consultations and clinical tests are obtained for this system development. Out of the 42 attributes obtained, 11 potential features were filtered with one of the Feature Engineering technique.  A PCOS grading system is performed using a variety of M L techniques, such as Decision Tree, Logistic regression, Naïve Bayes, K-Neighbour Neighbour (k-NN), Support Vector Machine (SVM) and Random Forest. Among all Logistic Regression proved to be most appropriate and accessible method.

Published

2020-11-01

Issue

Section

Articles