Feature level fine grained sentiment analysis using boosted long short-term memory with improvised local search whale...

Area of Science:

• Computational Linguistics
• Artificial Intelligence
• Machine Learning

Background:

• Online consumer reviews are crucial for e-commerce, providing insights into product perception.
• Sentiment analysis of these reviews faces challenges due to complex language and data variations.
• Accurate sentiment forecasting is vital for businesses and consumers in the digital marketplace.

Purpose of the Study:

• To propose a novel Improvised Local Search Whale Optimization improved Long Short-Term Memory (ILW-LSTM) model.
• To enhance feature-level sentiment analysis for online product reviews.
• To accurately classify sentiment into positive, negative, very positive, very negative, and neutral categories.

Main Methods:

• A multi-stage process involving data collection, pre-processing, and feature extraction using a modified inverse class frequency algorithm (LFMI).
• Feature selection is performed using the levy flight-based mayfly optimization algorithm (LFMO).
• Sentiment classification is achieved using the proposed ILW-LSTM model on the 'Prompt Cloud dataset'.

Main Results:

• The ILW-LSTM model achieved a high accuracy of 97% in sentiment classification.
• Performance evaluation using accuracy, recall, precision, and F1-score demonstrated the model's effectiveness.
• The proposed ILW-LSTM model significantly outperformed other leading algorithms in feature-level sentiment analysis.

Conclusions:

• The ILW-LSTM model offers a robust and accurate solution for feature-level sentiment analysis of online product reviews.
• The study highlights the effectiveness of integrating whale optimization and LSTM for complex sentiment classification tasks.
• This approach provides valuable insights for e-commerce platforms and consumer understanding.