I need to design a model by using python that can identify 3 different Arabic dialects by training and testing

my task is: I need to design a system or model by using python that can identify 3 different Arabic dialects (classes) in addition to MSA. by adopting a (1-3) word N-gram and (2-5) character N-gram to extract features to identify these dialects then will use the (MNB) Multinomial Naive Bayes classifier, in addition, also need to use Random Forest algorithm to select the top features later. for sure the model will be training first on these dialects and then will test it?
Notes: the model should be done by python and read from an excel file and for sure all the result will be saved in an excel file. for more information plz feel free to text me.