Topic Modeling of Papers related to Suicide Using Text Mining

Objectives: This research aimed to classify topics into published suicide-related papers and to understand the weight of major topics and the trend of changes in topics over the past 20 years. Methods: Articles were collected from KCI website and preprocessed. The revised data by preprocessing were analyzed using Latent Dirichlet Allocation (LDA) algorithm and major topics representing the articles were extracted. Topic trends by year and period were analyzed using the extracted topics. Results: Through the result of analyses, it was found that each topic has their own specific topics and they can be classified to their own topics rather than overlapping with other topics. We also found that the broadest topics are suicidal status, suicidal thoughts and suicidal prevention while elderly, youth, and risk groups are the target population for many researches. Conclusions: This research used big data and, through text mining, was able to find comprehensive trends. Using this research as a base, future researches on suicide and suicide prevention can analyze trends in order to study more specific group.