Abstract
COMPARATIVE STUDY OF MULTICOLLINEARITY USING REGULARIZATION METHOD
Ngozi Nzelu*
ABSTRACT
In this study, Ridge, Lasso and Elastic Net Regression were compared as a regularization method to determine the model that will be better to handle multicollinearity in a dataset, especially in the area of health. Variance inflation factors were used to dictate multicollinearity in liver patient record data set and and which correspond to Total Protiens (TP) and Albumin Ratio Albumin (ALB) respectively were highly correlated. The Statistical analysis result shows that the Elastic Net Regression performed better than Ridge and lasso Regression with minimum RMSE of 0.4311013 and highest R-square of 15.62376.
[Full Text Article] [Download Certificate]