Customer Experience towards the Product during a Coronavirus Outbreak

Area of Science:

• Natural Language Processing
• Machine Learning
• Consumer Behavior Analysis

Background:

• Sentiment analysis is crucial for marketing, guiding firms on consumer preferences and service improvement.
• Statistical methods can identify factors within consumer feedback, but advanced techniques are needed for nuanced analysis.

Purpose of the Study:

• To introduce a deep-based learning method, the deep sentiment classifier (DSC), for thorough product recommendation analysis.
• To investigate the effect sizes of positive, negative, and neutral feedback.
• To compare DSC performance against traditional machine learning classifiers.

Main Methods:

• Developed and applied a deep sentiment classifier (DSC), a deep-based learning method.
• Utilized a women's clothing review dataset (22,642 records) and the IMDB dataset (50,000 movie reviews).
• Performed 10-fold cross-validation and compared DSC against KNN, random forest, logistic regression, decision tree, SVM, multilayer perceptron, and naïve Bayes classifiers.

Main Results:

• Experimental studies indicate that recommendations are excellent positive sentiment indicators, outperforming fuzzy rating metrics.
• DSC achieved a top average F1 score of 93.56% for recommended form classification and 88.32% for recommended classification.
• DSC demonstrated superior performance compared to all tested machine learning classifiers on both datasets.

Conclusions:

• The deep sentiment classifier (DSC) provides a highly effective approach for sentiment analysis in consumer reviews.
• Recommendation status is a more reliable indicator of positive sentiment than numerical ratings.
• DSC offers significant improvements in sentiment classification accuracy for marketing and product review analysis.