Ridge and Lasso Regression for Feature Selection of Overlapping Ibuprofen and Paracetamol UV Spectra

S. Suprapto, Y. L. Nikmah


Feature selection is a process of identifying and selecting a subset of the input variables that are most relevant to the target variables. Regularized regressions such as Ridge and Lasso apply penalties to reduce the weight of features in the linear model input. Ridge regression causes the regression weight to be very close to zero, but not zero. On the other hand, the Lasso makes the weight equal to zero. Lasso regression is recommended for models with a lot of features, but only a few are important. In this study, a quantitative analysis of a mixture of ibuprofen and paracetamol in pain relief tablets was performed. A training set consisting of 25 UV spectra was optimized by Linear, Ridge, and Lasso regression. Two sets of test solutions with known concentrations of ibuprofen and paracetamol were used for validation. The Lasso regression only uses some absorbance values as estimators but has good accuracy at predicting ibuprofen and paracetamol in the test and pain relief tablet samples that were comparable to linear and ridge regression. The ridge regression showed better recovery and RMSE compared to the linear and lasso regression.


Regularized regression; feature selection; Ridge regression; Lasso regression; paracetamol; ibuprofen

Full Text:


DOI: https://doi.org/10.48317/IMIST.PRSM/morjchem-v11i1.31466