Core Concepts
The author compares Latent Dirichlet Allocation (LDA) and Non-negative Matrix Factorization (NMF) models to analyze aviation accident reports, highlighting the strengths of each model in uncovering latent themes and patterns within the data.
Abstract
This study delves into the analysis of aviation accident reports using topic modeling techniques, specifically LDA and NMF models. The research aims to automate the process of identifying latent themes within accident reports to enhance aviation safety measures. By comparing the performance of both models, the study provides insights into their effectiveness in extracting meaningful topics from a vast corpus of textual data. The findings suggest that while LDA demonstrates higher topic coherence, NMF excels in producing distinct and granular topics for a more focused analysis of specific aspects related to aviation accidents. Through rigorous methodologies for data preprocessing, model training, and evaluation based on coherence metrics, this research contributes to advancing safety protocols in the aviation industry.
Stats
LDA demonstrates higher topic coherence with a C_v score of 0.497.
NMF attained a C_v coherence score of 0.437.