American
International
University-
Bangladesh

Students & Supervisors

Abstract

The Cox Bazar Beach is the largest natural beach in the world that is highly overrun by tourists during key holidays. This short-lived and high-tide tourism causes an unreasonable burden on the local infrastructures, waste collection services and the wussy coastal ecologies. Conventional tools of a carrying capacity (that are static), based on predefined limits or qualitative judgments without consideration in the dynamic and non-linear interactions in the crowd environment. As long-term tourism is increasingly accessible, machine learning is an attractive alternative with predictive and adaptive analysis of carrying capacity which is a machinedriven methodology through the data. This research proposes a machine learning inspired method of the estimation of carrying capacity as well as crowd management planning at Cox Bazar during high festive seasons. The literature on tourism carrying capacity concentrates mainly on the physical, environmental and social thresholds by means of descriptive indicators-based approach, index based approach and system dynamics models. They are intuitive but of little prediction and correlation. Several work have employed machine learning for tourism demand forecasting as well as visitor flow prediction but few have focused on environmental stress prediction and overload detection. Another question that has not been studied is the dynamics of crowds in times of festivals, and explainable machine learning in the process of policy decision-making. These gaps are filled in this research through the combination of multi-model machine learning, pressure-based feature engineering, and explainability to study dynamic carrying capacity exceedance. A numerical data set was utilized between the period 1995-2024, and the parameters such as the total number of visitors, the highest number of visitors, hotel occupancy, waste production, and the parameters related to environmental stress were considered. New characteristics, such as visitor density ratio, waste produced per visitor, and pressure of occupancy were calculated to capture the impact of the crowd pressure in a better manner. The issue was formulated in a two-task typology and regression environment which incorporated machine learning methods. Five base models of machine learning were trained on a time conscious split between training and testing data to ensure that the integrity of time was preserved in the data set Linear Regression Model, Decision Trees Model, Random Forest Model, Gradient Boosting Model, and Support Vector Machine Model. Regression and classification measures were tested. The results reveal that the ensemble-based models which are the Random Forest and Gradient Boosting in particular are better than the linear and single-tree models in prediction and overload detection tasks. One of the outstanding performances is this nonlinearity of the interactions of the crowd and the environment were observed during the busiest holiday seasons. The results of the classification assist in identifying the importance of seasonal concentration and demonstrate high ability to determine the existence of overloads conditions despite the apparent manageability of the annual number of visitors. Peak season visitors and waste produced per visitor, followed by the pressure hotel occupancy is explainable factors in the causation of environmental stress. These findings prompt the need to have targeted crowd management, by establishing the fact that the issues of excess of the carrying capacity is largely due to festive crowd behavior as opposed to sustainable tourism development. Conclusion: In this research, a machine learning-based model of dynamic carrying capacity analysis and crowd control at the Bazar Beach in Cox is introduced during the periods of the busiest holidays. This proposed solution avoids the limitations of traditional methods of static analysis through a combination of explainable.

Keywords

Publication Details

• Type of Publication:
• Conference Name: RMSTU National Tourism Conference 2026
• Date of Conference: 16/01/2026 - 16/01/2026
• Venue: Rangamati Science and Technology University
• Organizer: Rangamati Science and Technology University