As cerebrovascular diseases, which are considered the main cause of death, often leave aftereffects once they occur, prevention is as important as treatment. For prevention purposes, studies are being conducted to establish the link between the frequency of treatments and meteorological and environmental factors for cerebrovascular diseases by applying a regression model or time series model. Social media buzz is an important source of insight into the rapid response of the public and is being used to predict information in various areas as social network usage increases.

In this paper, generalized linear model and generalized linear mixed model were established using weather, environment and social media information to predict the frequency of daily treatments for cerebrovascular diseases. Considering domain knowledge, we used the number of medical care inquired lagged 1 and 7 days, lowest temperature, lowest humidity, ozone lagged 1 day, carbon monoxide lagged 1 day, fine dust lagged 5 days and news buzz lagged 1 day as variables. Since the number of treatments is count data, model fitting was performed assuming that the Poisson distribution was followed. After fitting the model, various measures of assessment were used to perform the model evaluation and to compare the model through visualization. The results showed that a generalized linear mixed model with random slope for all categorical variables was the most appropriate model.