Public Perception Analysis of Tweets During the 2015 Measles Outbreak: Comparative Study Using Convolutional Neural...

Area of Science:

• Computational linguistics and public health informatics
• Application of deep learning for social media analysis in epidemiology

Background:

• Understanding public perception during health crises is crucial for effective public health responses.
• Social media platforms like Twitter offer real-time insights into large-scale public sentiment and reactions.

Observation:

• A comprehensive scheme was developed to analyze public perception of measles outbreaks from Twitter data, covering discussion themes, emotions, and vaccination attitudes.
• Over 1.1 million measles-related tweets were collected, with a gold standard set of 1151 tweets curated for analysis.
• Convolutional neural network (CNN) models were developed and compared against conventional machine learning methods using various word embedding configurations.

Findings:

• CNN models demonstrated superior performance in classifying measles outbreak-related tweets compared to traditional machine learning algorithms (k-NN, Naive Bayes, SVM, Random Forest).
• CNN models significantly improved recall, particularly for underrepresented classes in the highly unbalanced dataset.
• The best performing CNN model varied by classification dimension, with combined embeddings excelling in theme and emotion analysis, and Stanford embeddings in vaccination attitude analysis.

Implications:

• The developed scheme and CNN-based classification system can facilitate rapid, multi-dimensional analysis of public perception during infectious disease outbreaks.
• This approach offers a valuable tool for public health agencies to monitor and respond to public sentiment concerning diseases like measles, influenza, and Ebola.
• The study highlights the effectiveness of CNNs in handling small, imbalanced datasets common in real-world social media analysis for public health.