Fig. 3

Feature selection and cross-validation using LASSO. For different values of lambda, different numbers of features will be included in the final model and a decision has to be made. Figures A, C illustrate the process of cross validation in order to choose the most appropriate lambda value. Lambda resulting in the smallest mean squared error (MSE) is chosen, as the green lines shows. Figures B, D plot different values of lambda on the x-axis and each line represents a feature, showing when it enters the model and its level of influence on the outcome. A Selection of the tuning parameter (lambda) in the LASSO of unenhanced model. B The unenhanced CT feature coefficients varied according to lambda. C Selection of the tuning parameter (lambda) in the LASSO of enhanced model. D The enhanced CT feature coefficients varied according to lambda