Title

A Comparative Analysis of N-Gram Deep Neural Network Approach to Classifying Human Perception on Dengvaxia

Document Type

Article

Publication Date

2019

Abstract

The increasing use of social media platform like Twitter provides opportunity for information dissemination to the public. The Dengvaxia controversy in the Philippines negatively affected the public's perception towards vaccination. It has been noted that due to this incident, many parents have decided not to have their children vaccinated due to fear of endangering them [2]. This resulted to children contracting other diseases like measles due to the lack of immunization [1] [2]. The preference to not have newborns undergo vaccination program remains a threat to public health. Using publicly accessible tweets, this study aims to understand health perceptions of the public in relation to Dengvaxia. A deep neural network approach using n-gram vectorization is used in comparison to the Doc2Vec neural network classifier to identify tweets containing personal perception on health. It was discovered that not only does the bigram model perform better in classifying than the Doc2Vec model with a performance measure of 86.25% accuracy, 0.85 precision, 0.86 ROC and 0.85 F1 score, but also it is able to identify clearer and more diverse topic using LDA topic modeling in comparison with unigram and trigram model. This method allows the monitoring of public perception and acceptance towards the implementation of a new medication or vaccination especially after the Dengvaxia scandal that the Philippines experienced.

Share

COinS