A Floating Population Prediction Model in Travel Spots using Weather Big Data

Weather factors in travel spots and residential areas, such as temperature or precipitation, may cause a change in the floating population in travel spots during vacation seasons. This study aims to predict the daily floating population by creating a prediction model through a multiple linear regression analysis of the changes in the floating population based on weather factors. The regression analysis used 20 weather observation variables, 48 weather forecasting variables, and 6 dummy variables for the day. The three steps of the multiple linear regression analysis (creation of the exact model, removal of variable, and analysis of residuals) were performed to present the final multiple linear regression models for each of three famous travel spots in South Korea, Haeundae Beach, Gyeongpo Beach, and Daecheon Beach. The R square value of each model showed 6.2, 70.57, and 68.51% expression power. To verify the predictability, we evaluate the proposed model by comparing the predicted and real daily floating populations in July and August 2014. The evaluation method used MAPE, and the results showed 79.46, 65.2, and 65.94% accuracy levels, respectively.