A Study on the Estimation Model for the Visitors to Let’s Run Park Using Machine Learning

Abstract

PURPOSE The purpose of this study is to find the best model to predict the demand of visitors in Let’s Run Park by using machine learning and to provide effective data for establishing future marketing strategies. METHODS For this purpose, three methods of machine learning were applied: random forest, adaboost, and gradient boosting. The variables for predicting the audience were weather data and the number of visitors per date for four years as training data, and the accuracy was predicted by comparing the actual data for one year. RESULTS First, the performance evaluation using random forest was conducted, RMSE =1856.067, R2= .965, and error was 6.47%. Second, the performance evaluation using Adaboost was conducted, RMSE =1836.227, R2= .965, and error was 5.25%, which was the lowest among the three machine learnings. Third, the performance evaluation using gradient boosting showed that RMSE =1797.400 and R2= .967 were the most accurate among the three machine learnings and error was 6.99%. CONCLUSIONS As a result of this study, each of the three machine learning features existed, but the most efficient model was gradient boosting. In addition, the best way to utilize it in the field is to predict the number of visitors by comprehensively judging the results of the three machine learning, and it is judged that it will help efficient management decision making in the future.

Keywords

• demand forecasting model
• let’s run park(seoul)
• visitor
• machine learning