Exportar Publicação

Sequeiros, J.A., Ramos, F.R., Pereira, M.T. & Oliveira, M.J. (2022). Exploring the potential of Deep Learning to forecast purchases in the  Portuguese National Health Service. XXII Congresso da APDIO.

J. Sequeiros et al.,  "Exploring the potential of Deep Learning to forecast purchases in the  Portuguese National Health Service", in XXII Congr.o da APDIO, Évora, 2022

@misc{sequeiros2022_1766371018284,
	author = "Sequeiros, J.A. and Ramos, F.R. and Pereira, M.T. and Oliveira, M.J.",
	title = "Exploring the potential of Deep Learning to forecast purchases in the  Portuguese National Health Service",
	year = "2022",
	howpublished = "Digital",
	url = "http://apdio.pt/web/io2022"
}

TY  - CPAPER
TI  - Exploring the potential of Deep Learning to forecast purchases in the  Portuguese National Health Service
T2  - XXII Congresso da APDIO
AU  - Sequeiros, J.A.
AU  - Ramos, F.R.
AU  - Pereira, M.T.
AU  - Oliveira, M.J.
PY  - 2022
CY  - Évora
UR  - http://apdio.pt/web/io2022
AB  - The articulation of statistical techniques and tools, combining the mathematical and computational aspects, is manifested in clear support for decision-making, particularly in forecasting. Forecasting has been assumed as a fundamental tool in the creation of competitive advantage in any entity, contributing to the more efficient management of resources. 
The health sector is no exception, particularly in the financial management of health units. This fact gains even greater motivation and interest due to its importance for government entities as well as its impact on the general population. More efficient and effective management of logistics and supply chains is recognized across the board as one of the main areas of improvement, where the use of modeling and forecasting can be an excellent ally. 
To achieve this objective, methodologies of Artificial Neural Networks (ANN), namely Deep Neural Networks (DNN), have been pointed out in the scientific literature as a very promising option. In fact, in addition to the proposal of new ANN architectures, the combination of ANN with other methodologies has been frequently discussed in the literature.
Thus, starting from previous studies where analysis of historical data was made to forecast the charges in the Portuguese National Health Service (NHS), in this study we will try to evaluate the possibility and advantages of the implementation of Deep Learning methodologies. In addition to using classical methodologies (e.g., exponential smoothing models – ETS), DNN models and a hybrid model, BJ-DNN  are used. 
Comparative analysis is based on forecasting quality, considering Mean Absolute Percentage Error (MAPE) and computational cost associated with the forecast models used.
ER  -