Multivariate Nonlinear Regression with Artificial Neural Networks on Dynamic and Static Data
- Two data sets, one with sales data from Vodafone and another with a similar structure, were analyzed to predict a target variable. Five different models are compared with respect to the mean absolute error (MAE) of an independent test set. The models use a feedforward neural network (FFN), a long-term memory network (LSTM), and an auto-encoder network. The results show that a hybrid FFN-LSTM model performs best by reducing the MAE for the first data set to a value of 7.37, compared to 12.53 and 12.56 for other models.
| Author: | Seema Ganpat More, Rolf Bardeli, Martin Golz |
|---|---|
| URN: | urn:nbn:de:bsz:mit1-opus4-154246 |
| DOI: | https://doi.org/10.48446/opus-15424 |
| ISSN: | 1437-7624 |
| Parent Title (German): | 24. Nachwuchswissenschaftler:innenkonferenz |
| Subtitle (German): | Wissenschaftliche Berichte | Scientific reports |
| Publisher: | Hochschule Mittweida |
| Place of publication: | Mittweida |
| Document Type: | Conference Proceeding |
| Language: | English |
| Year of Completion: | 2024 |
| Publishing Institution: | Hochschule Mittweida |
| Contributing Corporation: | Hochschule Schmalkalden |
| Release Date: | 2024/07/11 |
| Tag: | Feedforward neural network; Long-term memory network; Mean absolute error |
| GND Keyword: | Feedforward-Netz |
| Issue: | 3 |
| Page Number: | 7 |
| First Page: | 333 |
| Last Page: | 339 |
| Open Access: | Frei zugänglich |
| Licence (German): |  Urheberrechtlich geschützt |
