Sentiment Analysis of Starbucks Reviews Using Machine Learning and Deep Learning
A recent research paper on arXiv (2608.12007) offers a comparative analysis of customer sentiments regarding Starbucks through both classical and deep learning techniques. The study utilized a dataset of over 700 reviews sourced from ConsumerAffairs, which underwent preprocessing to identify temporal and geographic trends. Sentiment labels were assigned based on star ratings, categorizing 4-5 stars as positive and 1-3 stars as negative, leading to a dataset skewed towards negative sentiments. The analysis compared five classical classifiers (Logistic Regression, SVM, Decision Tree, Random Forest, Naive Bayes) against five deep learning models (LSTM, RNN, Bidirectional LSTM, among others). This research seeks to enhance understanding of consumer sentiment in the coffee retail sector, emphasizing the role of reviews in influencing brand image and business tactics.
Key facts
- Study compares classical machine learning and deep learning for sentiment analysis of Starbucks reviews.
- Dataset from ConsumerAffairs contains over 700 reviews.
- Sentiment labels binarized from star ratings: 4-5 positive, 1-3 negative.
- Dataset imbalanced toward negative sentiment.
- Five classical classifiers evaluated: Logistic Regression, SVM, Decision Tree, Random Forest, Naive Bayes.
- Five deep learning models evaluated: LSTM, RNN, Bidirectional LSTM, and others.
- Preprocessing and exploratory data analysis identified temporal and geographic patterns.
- Research focuses on retail coffee sector and consumer reviews' impact on brand perception.
Entities
Institutions
- ConsumerAffairs
- Starbucks
- arXiv