A carregar...
Publicação
Unravelling the future: a hybrid approach to time series forecasting for customer support optimization at NOS
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.77 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
This work project compares individual and hybrid forecasting approaches for
efficient call center queue management at NOS. Besides simple individual models like theta
and more complex ones such as SVMs, hybridization methods, including linear parallel hybrid
and series hybrid combinations were tested. Statistical models, notably the theta model,
effectively capture key patterns and outperform more complex approaches. Meanwhile, a
parallel hybrid of Pareto-efficient models marginally improved performance was offset by
increased complexity. Overall, this work suggests that increasing complexity is unjustified, and
statistical methods can adequately forecast the task at hand.
Descrição
Palavras-chave
Forecasting Time series Hybrid combination modelling Data science