Sentiment analysis of pilgrims using CNN-LSTM deep learning approach.

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Computational Social Science

Background:

• Crowd management is crucial for large-scale events like the Hajj, the world's largest annual Muslim gathering.
• While technological monitoring of crowds is advancing, pilgrim sentiment and experiences remain under-researched, with limited social media data.
• Understanding pilgrim perspectives is vital for improving services and ensuring a meaningful experience.

Purpose of the Study:

• To develop and evaluate a hybrid deep learning model for sentiment analysis of Hajj pilgrims' experiences.
• To address the gap in understanding pilgrim opinions, particularly regarding services like catering.
• To create and utilize a specialized dataset, Catering-Hajj, for sentiment analysis.

Main Methods:

• A novel model combining Convolutional Neural Networks (CNN) for feature extraction and Long Short-Term Memory (LSTM) for dependency analysis was developed.
• The model employed a four-CNN layer architecture followed by an LSTM layer and a SoftMax classifier.
• A specialized dataset, Catering-Hajj, was collected, annotated, and used to analyze pilgrim sentiment towards pre-prepared food services during Hajj 1442.

Main Results:

• The proposed CNN-LSTM model achieved a 92% accuracy rate for sentiment analysis on the Catering-Hajj dataset.
• Performance was compared against various Machine Learning (ML) models, including SVM, LR, DT, RF, CNN, and LSTM.
• The hybrid model significantly outperformed existing methods, with SVM, RF, and LSTM achieving approximately 81% accuracy.

Conclusions:

• The developed CNN-LSTM model offers a highly effective approach for sentiment analysis in the context of large religious gatherings like Hajj.
• This research highlights the importance of analyzing pilgrim sentiment to enhance service quality and overall experience.
• The specialized Catering-Hajj dataset and the proposed model provide a valuable resource for future research in event management and computational social science.