This project is part of the project I had to conduct during my Msc at Ecole Polytechnique involving Capgemini consulting.
We were to construct a classifier to predict rating of Novotel Hotels. The big challenge in this case lied in the problem of unbalanced labels completely biasing the multinomial classifier we were to construct. I tried several methods to adjust this issue (oversampling, undersampling, automatically using SMOTE) whilst paying careful attention to when the <>sampling method was used as to not add bias to the model.
To train the classifier and test it, I constructed a pipeling to apply to a dataframe.