Performances of Machine Learning Models and Featurization Techniques on Amazon Fine Food Reviews
Rishabh Singh, Akarshan Kumar, Mousim Ray · 2023
In this technologically advanced world, the existence of online stores, which makes products available to the consumers on their fingertips, it is important for retailers to take feedback on their products by the consumers. Reviewing a product also gives other customers an idea about how good or bad a product could be. Reviews will help the retailers improve their service or their product. Our work aims to automatically analyze the reviews and classify if a review is good or bad using amazon fine food reviews dataset. To classify the reviews as positive or negative, four different kind of classifiers such as logistic regression, support vector machine (SVM), random forest and XGBoost were used. Each classifier was used along with four different vectors such as bag of words, TFIDF, average of word2vec, and a combination of TF-IDF and word2vec. The paper proposes a framework to automate the text analysis of reviews for polarization of its sentiment. The work also emphasizes that for certain datasets, simpler models and simpler featurization gives higher accuracy and performance compared to complex models.