Predicting and explaining customer satisfaction: A deep learning and sentiment analysis of emotional impacts

Area of Science:

• Business Analytics
• Computational Linguistics
• Consumer Psychology

Background:

• Customer satisfaction is crucial for retail success.
• Understanding emotional drivers enhances consumer experience and loyalty.
• Analyzing online reviews offers valuable insights into customer sentiment.

Purpose of the Study:

• To analyze customer satisfaction in retail using sentiment and emotion analysis of Yelp reviews.
• To identify key emotional drivers influencing customer satisfaction.
• To validate a keyword-based classification approach using deep learning models.

Main Methods:

• Utilized a dataset of over 500,000 Yelp reviews.
• Applied keyword-based feature engineering and ChatGPT for keyword generation.
• Employed deep learning models (LSTM, CNN) and NRC emotional lexicon.
• Conducted Ordinary Least Squares (OLS) regression analysis.

Main Results:

• Sentiment, trust, and joy positively correlate with customer satisfaction.
• Fear, surprise, anticipation, sadness, and disgust negatively impact customer satisfaction.
• Deep learning models validated the keyword-based approach.

Conclusions:

• An integrative approach combining deep learning and lexicon-based emotion analysis provides nuanced insights.
• Identified specific emotional drivers essential for improving retail consumer experience.
• The methodology offers a diagnostic framework for understanding customer satisfaction.