Lasso as a Sparse Modelling Technique is not a good choice in case of Mixed predictor-type problems. Instead, a combination of Group Lasso, Variable Fusion, and Fused Lasso Penalty should be used. A Data Analysis is performed to check the adequacy of the suggested method.
Original Paper : Sparse modeling of categorial explanatory variables
However, we have used a more general method called the "smurf Algorithm". Its implementation is available in R as Package - smurf: Sparse Multi-Type Regularized Feature Modeling