Analysis of Factors Affecting Tourist Spot Revisitation After Natural Disaster Calamity: A Machine Learning Ensemble Approach

Abstract

The tourism industry provides the Philippines with trillions of annual monetary incomes. However, being a country prone to natural disasters, many tourist attractions or destinations are damaged, which causes tourists to become hesitant. This study aimed to assess factors influencing tourist revisitation behavior towards disaster-stricken attractions by utilizing an extended version of the Theory of Planned Behavior, and by predicting human behavior through the Machine Learning Algorithms (MLA) such as Decision Tree (DT), Random Forest Classifier (RFC) and Artificial Neural Network (ANN). Through the use of MLA, this study was able to provide justification of its accurate results in identifying significant latent variables. From this study's findings, feature selection using correlation analysis was a good technique for data pre-processing. Moreover, it was proven that ANN help provided better insights from the RFC results. In addition, RFC overpowered the basic decision tree, having significant differences in the output. The discussion section is arranged from the most to the least important variable. A total of 1,008 Filipinos voluntarily answered an online questionnaire that consisted of 45 questions, leading to 45,360 datasets. With 97.86% and 96% accuracy rates for ANN and RFC, respectively, hedonic motivation was the most important factor affecting tourists' revisitation behavior. It may also be posited that the other significant contributing factors are intention, perceived behavioral control, and media. Interestingly, tourists' hedonic motivation relatively outweighed their disaster concern, and DT did not achieve accurate results as it only had an undesirable 56.97% accuracy rate. Finally, the managerial insights provided in this study could be applied and extended to tourism industries in different countries. Likewise, the MLA and its implications seen in the study may be considered for predicting other behavioral intention studies that aim to assess tourism for other tourist attractions.