Ispany , MartonAung, Htet Htet2024-02-012024-02-012023-11-15https://hdl.handle.net/2437/365969This study has applied five different machine learning algorithms of Bernoulli Naive Bayes, Complement Naive Bayes, Logistic Regression, Linear Support Vector Machine and Random Forest on the Amazon fine food products reviews to predict sentiment (positive or negative). As mentioned above, the dataset is imbalanced. Therefore, the data has been prepared to be balanced and qualified using random resampling and random undersampling. After balancing the data, the size of data has been decreased due to undersampling. This study has conducted three experiments using unigrams, bigrams, and trigrams in TF-IDF. The results from the study showed that in terms of accuracy, the Linear Support Vector Classifier model got the highest accuracy (91.21) and outperformed among the proposed models while the Naive Bayes models had the lowest accuracies. In terms of running time, Naive Bayes models were the fastest ones than the other models while the Random Forest Classifier consumed the longest time among other models.41enNatural Language Processing (NLP),Sentiment AnalysisText MiningMachine LearningTokenizationWord2vecN-gramProduct ReviewsDEENK Témalista::InformatikaHozzáférhető a 2022 decemberi felsőoktatási törvénymódosítás értelmében.