Churn Prediction Model in Healthcare

This experiment is done to predict whether member gets enrolled or not in the subsequent years with historical data of claims . This will identify top reasons for churn in order to improve health insurance coverage. What data do we need ? Enrollments,Plan,Member demographics ,Claim,Call tracking information ,grievances from Payer Databases Based on a research top reasons could be Rise in Net Premium than previous years Forget to sign up Auto renewal set to No Move to Medicaid /Public coverage Income might change every year Moving from one state to another Maximum Denial of Claims Maximum grievances raised by Member Certain life events It is quite surprising that everything has a pattern. Some patterns we understand and others we do not as of now. So in broader terms, we will be able to predict future for events with pattern . Since we are going to predict enrollment status for subsequent year ,Enrollment status is the target variable which will be predicted as Yes /No . For such problems we have to choose classification ML algorithm Two Class Logistics Regression Two Class Decision Forest AUC -Area under the curve shows the model accuracy--- more area you get more accurate are the results -shows 97 % Accuracy shows 91% Business Impact Predictions from machine learning can make apps and devices smarter. Predictive Analytics for financial success Attract new Customers having such new digital features in legacy products. Increase in sales Increase in revenue