This study aims how to find the spatial characteristics by region on the stress sentiment and the topics extracted from tweet data by twitter users. We downloaded 503,737 tweet data from July 12, 2014 to July 21, 2015, but analyzed 332,328 tweet data excluding advertisement and news. We classified tweet texts into morphemes in order to analyze the tweet data and applied the topic modeling based on LDA algorithm to search the topic about stress. As a result, 15 topics were generated and classified into 3 categories such as causes, results and relief methods of stress. The topics related to causes of stress are “personality”, “learning”, “job”, “family”, and “SNS use”, those related to results of stress are “illness”, “mental status”, “hair loss” and those related to relief methods of stress are “painting”, “gaming”, “exercise and cultural life”, “food intake”, “music”. Tweet data that home location are clearly identified are only 1,035 among 332,328 tweet data. So, we geolocated 34,641 tweet data through the residence model. We analyzed the results of the topic modeling by using 17 administrative districts (Si-Do) and then identified maximum and minimum number of topics. Finally, we compared which topics are highly interested in 17 administrative districts according to causes, results, and relief methods of stress. This paper is meaningful in two folds: First, this paper shows how to represent the meaning of people's sentiment which is the social pathology of stress. Second, this paper extends the field of spatial data by mapping invisible and emotional phenomena.