Comparison of SVM and Naive Bayes for Sentiment Classification using BERT data

Expressing views related to any common interest of area is becoming very general due to the excessive use of internet on the handheld devices like mobiles. Everyone is free to share his opinions within his social network. Having some useful knowledge from those opinions is the challenge for text classification techniques. In this work, the movies related opinions in the form of long and more than one sentence are analyzed to categorize them as positive or negative sentiments towards the movie. In this paper, the machine learning approach using the supervised learning methods is used for the implementation of the model. The data is prepared using the BERT model where plain text is directly converted into a numerical data form. The experiments are conducted using the Support Vector Machine and Naive Bayes algorithms. Accuracy, F1 score, and other metrics have been evaluated for the comparison of the results of both algorithms. The results show that the predictions by SVM have high accuracy than Naive Bayes.