Project of Machine Learning course, Data Science and Business Informatics - University of Pisa
A. Carnevale, F. Canepuzzi, G. Segurini
For this project it is required to solve a regression task on the CUP dataset. It is focused on trying out the different models, selecting the best hyperparameters configuration and compare their performances. We used: KNN, SVM, LBE, Random Forest and Neural Network. These models were first used to perform a classification task on the well-known MONK dataset as a benchmark. Finally, we implemented from scratch a Stacking and a Voting ensemble by combining the tuned estimators above.
Performance of the models evaluated with a Crossvalidated approach
Performance of the ensembles using as base estimators a subset of the models above