An analysis of Disaster Risk Suggestions using Latent Dirichlet Allocation and Hierarchical Dirichlet Process (Nonparametric LDA)

Abstract

Qualitative data is part of the things that most social scientists would deal with. In this study, qualitative Disaster Risk Reduction suggestions were analyzed using topic modeling techniques. Latent Dirichlet allocation is one of the topic modeling that was utilized in this study. The ideal number of topic models being generated for LDA is 10 with a score of 530.1495. Hierarchical Dirichlet Process model was also used to get the topic models from the corpus. The HDP model generated 11 topic models with a log-likelihood score of -4.08997. The topic models being generated by the parametric LDA and non-parametric LDA are almost similar. To analyze the result of the topic models, open coding technique was utilized. The following narratives were the focus of the DRR responses: Solid waste management and improve drainage system, Relief and Emergency Plan and Early warning system and Disaster Preparedness.