DOI: https://doi.org/10.59321/BAUETJ.V4I2.13
AUTHOR(S)
Kakoly Islam Tanjum1*, Tawhid Ahmed Komol1, Sheak Rashed Haider Noori1
ABSTRACT
Predicting customer sentiment from social media Bengali food reviews using a combination of traditional machine learning and deep learning algorithms. The dataset includes customer reviews labeled ‘positive’ or ‘negative,’ showing views on various food experiences. To analyze and classify the sentiment of Bengali food reviews, multiple algorithms, such as Bernoulli Naive Bayes, Support Vector Machine, Logistic Regression, K-Nearest Neighbors, Decision Tree Classifier, Long Short-Term Memory, and Convolutional Neural Networks are used. The study started by creating an overview for sentiment classification using classic machine learning algorithms such as BNB, SVM, LR, KNN, and Decision Tree. Following that, deep learning models such as LSTM and CNN are used to harness neural network power in collecting complicated patterns in text-based information. CNN outperforms all other algorithms, achieving an impressive accuracy of 97.60% in predicting customer sentiment. CNN performs better than other models because it can learn organized visualizations of features. This ability helps it identify specific context information and small differences found in Bengali food reviews.