Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neural Network Modeling for Forecasting Tourism Demand in Stopića Cave: A Serbian Cave Tourism Study

Buda Bajić, Srđan Milićević, Aleksandar Antić, Slobodan Marković, Nemanja Tomić

TL;DR

This paper addresses forecasting monthly tourism demand for Stopića cave in Serbia using ARIMA, SVR, and NeuralProphet. It demonstrates that NeuralProphet, a hybrid model that blends classical time-series components with neural networks and incorporates Google Trends as an exogenous covariate, provides the best predictive accuracy on a 12-month ahead horizon. Incorporating seasonality, trend, nonlinearity, and exogenous inputs yields substantial improvements over pure ARIMA or SVR and yields interpretable parameter outputs for decision-makers. The findings have practical implications for carrying-capacity planning and ongoing cave monitoring to support sustainable geotourism.

Abstract

For modeling the number of visits in Stopića cave (Serbia) we consider the classical Auto-regressive Integrated Moving Average (ARIMA) model, Machine Learning (ML) method Support Vector Regression (SVR), and hybrid NeuralPropeth method which combines classical and ML concepts. The most accurate predictions were obtained with NeuralPropeth which includes the seasonal component and growing trend of time-series. In addition, non-linearity is modeled by shallow Neural Network (NN), and Google Trend is incorporated as an exogenous variable. Modeling tourist demand represents great importance for management structures and decision-makers due to its applicability in establishing sustainable tourism utilization strategies in environmentally vulnerable destinations such as caves. The data provided insights into the tourist demand in Stopića cave and preliminary data for addressing the issues of carrying capacity within the most visited cave in Serbia.

Neural Network Modeling for Forecasting Tourism Demand in Stopića Cave: A Serbian Cave Tourism Study

TL;DR

This paper addresses forecasting monthly tourism demand for Stopića cave in Serbia using ARIMA, SVR, and NeuralProphet. It demonstrates that NeuralProphet, a hybrid model that blends classical time-series components with neural networks and incorporates Google Trends as an exogenous covariate, provides the best predictive accuracy on a 12-month ahead horizon. Incorporating seasonality, trend, nonlinearity, and exogenous inputs yields substantial improvements over pure ARIMA or SVR and yields interpretable parameter outputs for decision-makers. The findings have practical implications for carrying-capacity planning and ongoing cave monitoring to support sustainable geotourism.

Abstract

For modeling the number of visits in Stopića cave (Serbia) we consider the classical Auto-regressive Integrated Moving Average (ARIMA) model, Machine Learning (ML) method Support Vector Regression (SVR), and hybrid NeuralPropeth method which combines classical and ML concepts. The most accurate predictions were obtained with NeuralPropeth which includes the seasonal component and growing trend of time-series. In addition, non-linearity is modeled by shallow Neural Network (NN), and Google Trend is incorporated as an exogenous variable. Modeling tourist demand represents great importance for management structures and decision-makers due to its applicability in establishing sustainable tourism utilization strategies in environmentally vulnerable destinations such as caves. The data provided insights into the tourist demand in Stopića cave and preliminary data for addressing the issues of carrying capacity within the most visited cave in Serbia.
Paper Structure (14 sections, 9 equations, 7 figures)

This paper contains 14 sections, 9 equations, 7 figures.

Table of Contents

  1. Introduction
  2. Theoretical background
  3. Method
  4. ARIMA
  5. SVR
  6. NeuralPropeth
  7. Evaluation and Results
  8. Data description and experiment design
  9. Evaluation
  10. Results
  11. Discussion
  12. Insights into Stopića cave tourism demand
  13. Tourism demand and cave monitoring
  14. Conclusion

Figures (7)

  • Figure 1: SVR model.
  • Figure 2: Number of Stopića cave visitors during 2010-2023, monthly frequency.
  • Figure 3: Number of Stopića cave visitors during 2010-2023 (scaled to $[0,100]$) and Google Trend index.
  • Figure 4: Prediction of number of visitors for 12 months for 2023 obtained with different versions of ARIMA model. The most accurate predictions gave SARIMAX - ARIMA which includes a seasonal component and Google Trend as an exogenous variable.
  • Figure 5: Prediction of number of visitors for 12 months for 2023 obtained with SVR model, $\text{RMSE} = 4430.33$.
  • ...and 2 more figures