Leveraging hybrid model for accurate sentiment analysis of Twitter data

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Social Media Analytics

Background:

• Sentiment analysis is crucial for understanding public opinion in digital environments.
• Twitter is a key platform for real-time social media engagement and data collection.

Purpose of the Study:

• To analyze sentiments from Twitter data using advanced artificial intelligence techniques.
• To evaluate the effectiveness of a combined Bi-LSTM and Logistic Regression model for sentiment classification.

Main Methods:

• Data pre-processing using natural language processing (NLP) techniques: tokenization, stop-word removal, and stemming.
• Feature representation using Bi-Directional Long Short-Term Memory (Bi-LSTM) networks to capture sequential patterns.
• Sentiment classification using a Logistic Regression model with optimized hyperparameters.

Main Results:

• The integrated approach achieved high performance metrics: 81.8% precision, 83.4% recall, 82.5% F1-score, and 82.32% accuracy.
• Demonstrated the efficacy of combining Bi-LSTM for feature extraction and Logistic Regression for classification.
• Successfully classified sentiments as positive or negative from unstructured Twitter data.

Conclusions:

• The combined Bi-LSTM and Logistic Regression model provides a robust framework for sentiment analysis on social media.
• This approach shows significant potential for enhancing sentiment classification tasks in dynamic digital landscapes.
• Highlights the power of advanced AI and NLP in analyzing large-scale textual data for public opinion insights.