F1group_abstract.txt

The final model we decided to deliver is an ensembling of the following models: a neural network model implemented with PyTorch, a neural network model implemented with Keras, a K-nearest regression model implemented with scikit-learn, a K-nearest Regression model implemented with scikit-learn, a Kernel Ridge Regression model implemented with scikit-learn which gives in output only the predicted labels for the first target and a Support Vector Regression model implemented with scikit-learn which gives in output only the predicted labels for the second target. For the model selection we always performed a cross validation, while the model assessment phase was performed through a hold out, using a test set composed of some patterns exclusively selected for this aim from the original dataset ML-CUP22-TR.csv.