EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

A Comparative Evaluation of Univariate Time Series Forecasting Models for Tourism Operational Indicators at Boom Marina Banyuwangi

Muhammad Ilham Choiruddin, Ego Widiarto

Journal of Artificial Intelligence and Data Science · 15.12.2025 · 1 (2) · 66-74

DOI: 10.67561/jaids.v1i2.2 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Artificial Intelligence and Data Science, online erschienen am 15.12.2025. Verfasst von Muhammad Ilham Choiruddin und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate forecasting of a tourism operational indicator is crucial for planning and making informed decisions, but heterogeneity in temporal patterns prevents us from selecting a single optimal model. This research aims to compare univariate time-series forecasting models to identify the best model for each operational index in the Boom Marina Banyuwangi Tourism Area. Data on weekly ticketing_log from PT Pelindo Properti Indonesia covering seven input indicators were used, namely the number of tourists, total income, the number of motorcycles, the number of vehicles, the number of pedestrians, the cash payment amount, and the total parking fee. The flow of this study follows the CRISP-DM framework. It includes (i) data cleansing (using Z-score for winsorization), median imputation, daily-to-weekly aggregation, Robust Scaler normalization, (ii) sliding window formation (4, 8, 12 weeks), (iii) separation into training and test set with a ratio of 80:20. Seven models were used for comparison (BiLSTM, BiGRU, LSTM, GRU, Simple RNN, 1D-CNN, and a Random Forest Regressor) which performance was measured in terms of MAE or RMSE. The findings demonstrate that no single model performed best across all measures. Random Forest Regressor was the best predictor of four indicators: number of tourists (MAE = 0.5053; RMSE = 0.6532) and number of vehicles (MAE = 0.5904; RMSE = 0.7966). In contrast, BiGRU, BiLSTM, and Simple RNN achieved the best performance for total revenue, the number of motorcycles, and the number of pedestrians, respectively. These results imply that tree-based ensemble methods can capture nonlinear patterns, whereas recurrent neural networks are well-suited for indicators with strong temporal dependencies.

Zitieren

APA

Choiruddin, M. I., & Widiarto, E. (2025). A Comparative Evaluation of Univariate Time Series Forecasting Models for Tourism Operational Indicators at Boom Marina Banyuwangi. Journal of Artificial Intelligence and Data Science, 1(2), 66-74. https://doi.org/10.67561/jaids.v1i2.2

Vancouver

Choiruddin MI, Widiarto E. A Comparative Evaluation of Univariate Time Series Forecasting Models for Tourism Operational Indicators at Boom Marina Banyuwangi. Journal of Artificial Intelligence and Data Science. 2025;1(2):66-74. doi:10.67561/jaids.v1i2.2

BibTeX
@article{Choiruddin2025,
  author = {Choiruddin, Muhammad Ilham and Widiarto, Ego},
  title = {A Comparative Evaluation of Univariate Time Series Forecasting Models for Tourism Operational Indicators at Boom Marina Banyuwangi},
  journal = {Journal of Artificial Intelligence and Data Science},
  year = {2025},
  volume = {1},
  number = {2},
  pages = {66--74},
  doi = {10.67561/jaids.v1i2.2}
}
Kostenlos starten →