Designing Intelligent Models with ARIMAANN for Visionary Forecasts

Abstract

It is essential to choose the right model that can explain the growth of tourism in Albania and, therefore, to make the right decisions and direct the flow of tourists. This research aims to compare and apply three forecasting models: ARIMA, Artificial Neural Networks (ANN), and the hibrid ARIMAANN model to forecast the number of international tourists in Albania. The outcomes indicate that the interaction between the two approaches, ARIMAANN, is the first model to explain 96% of the data variation and provides the minimum mean absolute percentage error (MAPE) of 21.6%. In order to enhance the model's precision, the refined model, ARIMAANN 21-24, was suggested, which excluded the pre-pandemic and pandemic periods. This adjustment resulted in significant enhancements where the accuracy was 0.99, and MAPE was 7.09 %, making it the most accurate forecast. The proposed model shows that tourism will keep increasing in the next five years. The most tourists are expected in August 2029, with 2.9 million international tourists. This research provides a predictive tool for policymakers, tourism operators, and government agencies to capitalize on the benefits of hybrid modelling to enhance sustainable tourism's strategic development and management. All data and analyses were processed in RStudio with the latest advancements in time series modelling.