LSTM-DGWO-Based Sentiment Analysis Framework for Analyzing Online Customer Reviews

Area of Science:

• Natural Language Processing
• Machine Learning
• Deep Learning

Background:

• Sentiment analysis is crucial for product enhancement but traditional machine learning methods are computationally intensive and unreliable.
• Deep learning models like Long Short-Term Memory (LSTM) show promise, yet hyperparameter optimization remains a challenge.

Purpose of the Study:

• To develop a more accurate and efficient sentiment analysis model for app reviews.
• To address the hyperparameter optimization issue in deep learning models for sentiment analysis.

Main Methods:

• A hybrid model combining Long Short-Term Memory (LSTM) with Differential Grey Wolf Optimization (DGWO) was proposed.
• Bidirectional Encoder Representations from Transformers (BERT) were used for efficient word embeddings.
• Genetic Algorithm (GA) and Firefly Algorithm (FA) were employed for feature extraction and selection.

Main Results:

• The proposed LSTM-DGWO model achieved a high accuracy of 98.89% in categorizing app reviews.
• The model demonstrated superior performance compared to conventional sentiment analysis methods.

Conclusions:

• The study successfully developed an advanced sentiment analysis model for practical application in understanding customer perception.
• The findings highlight the potential of the proposed model for product enhancement from a business perspective.