In this paper, human location information is regressed by machine learning to predict the visiting ratio for a specific location category. The location data collected by the participants using the smartphone. Personal factors obtained through the ques...
In this paper, human location information is regressed by machine learning to predict the visiting ratio for a specific location category. The location data collected by the participants using the smartphone. Personal factors obtained through the questionnaire. We analyzed location data and personal factors by three ensemble techniques of machine learning: random forest, XGBoost, and stacking. A total of 34 participants collected their location data for 3 to 6 months. Before the learning of the machine learning model, the feature selection process was performed, and then the features influencing the results were extracted and tested. In addition, Grid search, Random search, and Bayesian optimization are used to optimize the hyperparameters that affect the performance of the ensemble model. As a result, it was confirmed that the performance of the model can be improved through optimization of hyperparameters. Also, meaningful prediction accuracy values were obtained, except for the two location categories, in which data were insufficient due to the fact that the experiment participants rarely visited. These results were revealed through three machine learning techniques to increase the reliability of the results.