4. A Comparative Analysis of The Performance of Some Penalized Regression Techniques in The Presence of Multicollinearity
Authors- Nwuzor, Ozoemena, A.U .UDOM
Abstract- Multicollinearity is a common issue faced by statisticians and machine learning practitioners when building predictive models. This study compared the performance of some penalized re-gression techniques (Lasso, Ridge and Elastic Net) in the presence of multicollinearity using real-life datasets. The comparison was carried out in terms of the accuracy, precision, and recall scores of each technique using the root mean square error, residual sum of squares and R-square. The outcome of this study using the real-life dataset on 442 diabetes patients measured on 10 baseline predictor variables and one measure of disease progression showed that the Ridge regression performed better than Lasso and Elastic net. Comparing the results of the methods, It was observed that Ridge model performed better in comparing the performance of some penalized regression models with multicollinearity.
