This paper classifies sentiment analysis in Arabic language and mining sentiment in relation to the COVID-19 pandemic in the period (2019 - 2021). Three large data sets are collected from tweets, hotel and restaurant reviews for building the proposed sentiment analysis model. We compared eight machine learning algorithms ,Multinomial Naïve Bayes (MNB), Bernoulli Naïve Bayes (BNB), Decision Tree (DT) ,K-nearest neighbor classifier (KNN), Support Vector Machines (SVM), Linear Support Vector Classifier (LSVC), Random Forest Classifier (RFC) and Stochastic Gradient Descent Classifier (SGD) on three cases: n-gram unigram, bigram, and trigram for each algorithm. The performance evaluations are compared according to precision, recall, and F-measure. The polarity prediction results in sentiment analysis models was achieved by linear SVC using hotel datasets with bigram case , with the accuracy of 0.966, precision of 0.967, recall of 0.966 and F-measure of 0.966 . The rest algorithms give average performance on all datasets . It can be concluded that the machine learning algorithms need the right morphological features to enhance the classification accuracy when dealing with different words that play different roles in the sentence with the same letters.