Sentiment analysis is a process during which the semantic orientation or polarity (i.e. positive, negative or neutral) of a given text is determined. This work deals with the sentiment analysis for Facebook's comments written in Arabic Modern Standard or Moroccan Dialectal from a Machine Learning perspective. The process starts by collecting and preparing the Arabic Facebook comments that we have annotated using crowdsourcing. Then, several combinations of extraction and weighting schemes for features construction was conducted to ensure the highest performance of the developed classification models. In addition, to reduce the dimensionality and improve the classification performance, a features selection method is applied. Our Machine Learning approach was implemented with the purpose of analysing the Facebook comments, written in Modern Standard Arabic or in Moroccan Dialectal Arabic, on the real data.