Exploring Female Infertility Using Predictive Analytic

Abstract

With the availability of medical data for large number of patients in hospitals, early detection of diseases has been made easier in the recent past. Conditions like Infertility which are hard to detect or diagnose can be now diagnosed with greater precision with the help of predictive modeling. One of the key challenges for early detection and timely treatment is in identifying and recording key variables that contribute to specific variance of infertility. In this paper, we consider 26 variables and identify relevant variables for early detection of 8 variant classes of female infertility. We compared various techniques and determined that the Random forest is the best method offerings 88% of accuracy for a reasonably large hospital dataset of size 965.