Artificial intelligence

Abstract: This paper investigates the applicability of the global neural network model N-BEATS for forecasting demand for 14 automotive brands based on search activity data for 2023-2025, without exogenous factors. The model is trained simultaneously on all time series and evaluated over a 12-month horizon (2026); forecast accuracy is validated against actual data for January-May 2026. The model achieved an overall MAPE of 18.2% (RMSE = 10,136.7), outperforming the Seasonal Naive baseline (MAPE = 21.5%). For 9 of 14 brands, the global model proved more accurate than the baseline, with the strongest results for Suzuki, Mercedes-Benz, and Hyundai. Expected demand dynamics for 2026 relative to actual 2025 figures are estimated for each brand individually. The findings confirm that search activity serves as a viable leading indicator of demand and that global neural network models remain practical even with a limited length of individual time series.

Keywords: N-BEATS; global forecasting models; search activity; automobile demand; time series; neural networks; Google Ads.

References:

1. Oreshkin, B. N., Carpov, D., Chapados, N., & Bengio, Y. (2020). N-BEATS: Neural basis expansion analysis for interpretable time series forecasting. International Conference on Learning Representations (ICLR).
2. Oreshkin, B. N., Dudek, G., Pełka, P., & Turkina, E. (2021). N-BEATS neural network for mid-term electricity load forecasting. Applied Energy, 293, 116918. https://doi.org/10.1016/j.apenergy.2021.116918
3. Olivares, K. G., Challu, C., Marcjasz, G., Weron, R., & Dubrawski, A. (2023). Neural basis expansion analysis with exogenous variables: Forecasting electricity prices with NBEATSx. International Journal of Forecasting, 39(2), 884-900. https://doi.org/10.1016/j.ijforecast.2022.03.001
4. Montero-Manso, P., & Hyndman, R. J. (2021). Principles and algorithms for forecasting groups of time series: Locality and globality. International Journal of Forecasting, 37(4), 1632-1653.
5. Zhang, M., Xu, H., Ma, N., & Pan, X. (2022). Intelligent vehicle sales prediction based on online public opinion and online search index. Sustainability, 14(16), 10344. https://doi.org/10.3390/su141610344
6. Ivashchenko, H. S., Tymoshenko, D. O., Blyzniuk, O. V., & Kononenko, O. M. (2024). Modeli hlybokoho navchannia dlia prohnozuvannia chasovykh riadiv [Deep learning models for time series forecasting]. Systemy upravlinnia, navihatsii ta zviazku. Zbirnyk naukovykh prats, 1(75), 82-87. https://doi.org/10.26906/SUNZ.2024.1.082
7. Ivashchenko, H. S., Ponamarov, V. V., & Kholiev, V. V. (2023). Korotkostrokove prohnozuvan-nia nestatsionarnykh chasovykh riadiv z vykorys-tanniam modelei MLP ta LSTM [Short-term forecasting of non-stationary time series using MLP and LSTM models]. Systemy upravlinnia, navihatsii ta zviazku. Zbirnyk naukovykh prats, 1(71), 91-95. https://doi.org/10.26906/SUNZ.2023.1.091
8. Abbasimehr, H., & Noshad, A. (2025). Localized global time series forecasting models using evolutionary neighbor-aided deep clustering method. Journal of Forecasting. https://doi.org/10.1002/for.3263
9. Kucher, P., & Yunkova, O. (2023). Prohnozuvannia dynamiky rynku vitaminiv za dopomohoiu neiromerezh [Forecasting vitamin market dynamics using neural networks]. Nauka i tekhnika sohodni, 3(17), 110–121. https://doi.org/10.52058/2786-6025-2023-3(17)-110-121
10. Kucher, P. V., & Yunkova, O. O. (2024). Modeliuvannia dynamiky rynku preparativ dlia volossia v umovakh pandemii COVID-19 [Modelling hair product market dynamics during the COVID-19 pandemic]. Problemy suchasnykh transformatsii. Seriia: ekonomika ta upravlinnia, 13. https://doi.org/10.54929/2786-5738-2024-13-11-01
11. Google LLC. (n.d.). Google Keyword Planner. Google Ads. https://ads.google.com/home/tools/keyword-planner/
12. Fantazzini, D., & Toktamysova, Z. (2015). Forecasting German car sales using Google data and multivariate models. International Journal of Production Economics, 170, 97–135. https://doi.org/10.1016/j.ijpe.2015.09.010
13. Abbasimehr, H., & Noshad, A. (2025). Localized global time series forecasting models using evolutionary neighbor-aided deep clustering method. Journal of Forecasting. https://doi.org/10.1002/for.3263