An Enhanced IDBO-CNN-BiLSTM Model for Sentiment Analysis of Natural Disaster Tweets

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Disaster Management

Background:

• Social media is crucial for disaster information dissemination.
• Accurate sentiment analysis of public tweets aids emergency response.
• Existing sentiment analysis models have limitations in applicability.

Purpose of the Study:

• To propose an enhanced IDBO-CNN-BiLSTM model for accurate emotional polarity recognition in disaster-related tweets.
• To improve upon existing swarm intelligence and deep learning models for sentiment analysis.
• To aid government and rescue organizations in understanding public demands during natural disasters.

Main Methods:

• Developed an improved Dung Beetle Optimization (IDBO) algorithm incorporating Latin hypercube sampling, Osprey Optimization Algorithm (OOA), and adaptive Gaussian-Cauchy mixture mutation.
• Utilized the IDBO algorithm to optimize hyperparameters for a Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) model.
• Applied the IDBO-CNN-BiLSTM model to classify emotional tendencies in tweets related to Hurricane Harvey.

Main Results:

• The IDBO-CNN-BiLSTM model achieved an accuracy of 0.8033 in classifying tweet emotional tendencies.
• The proposed model outperformed other single and hybrid models in empirical analysis.
• Accuracy was enhanced by 2.89% (vs. GWO), 2.82% (vs. WOA), and 2.72% (vs. DBO).

Conclusions:

• The IDBO-CNN-BiLSTM model demonstrates superior performance for sentiment analysis in disaster contexts.
• This model can effectively assist emergency decision-making during natural disasters.
• The research highlights the potential of hybrid AI approaches in disaster management and public sentiment understanding.