Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper › peer-review
Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper › peer-review
}
TY - GEN
T1 - An ensemble of single multiplicative neuron models for probabilistic prediction
AU - Yolcu, Ufuk
AU - Jin, Yaochu
AU - Egrioglu, Erol
PY - 2017/2/13
Y1 - 2017/2/13
N2 - Inference systems basically aim to provide and present the knowledge (outputs) that decision-makers can take advantage of in their decision-making process. Nowadays one of the most commonly used inference systems for time series prediction is the computational inference system based on artificial neural networks. Although they have the ability of handling uncertainties and are capable of solving real life problems, neural networks have interpretability issues with regard to their outputs. For example, the outputs of neural networks that are difficult to interpret compared to statistical inference systems' outputs that involve a confidence interval and probabilities about possible values of predictions on top of the point estimations. In this study, an ensemble of single multiplicative neuron models based on bootstrap technique has been proposed to get probabilistic predictions. The main difference of the proposed ensemble model compared to conventional neural network models is that it is capable of getting a bootstrap confidence interval and probabilities of predictions. The performance of the proposed model is demonstrated on different time series. The obtained results show that the proposed ensemble model has a superior prediction performance in addition to having outputs that are more interpretable and applicable to probabilistic evaluations than conventional neural networks.
AB - Inference systems basically aim to provide and present the knowledge (outputs) that decision-makers can take advantage of in their decision-making process. Nowadays one of the most commonly used inference systems for time series prediction is the computational inference system based on artificial neural networks. Although they have the ability of handling uncertainties and are capable of solving real life problems, neural networks have interpretability issues with regard to their outputs. For example, the outputs of neural networks that are difficult to interpret compared to statistical inference systems' outputs that involve a confidence interval and probabilities about possible values of predictions on top of the point estimations. In this study, an ensemble of single multiplicative neuron models based on bootstrap technique has been proposed to get probabilistic predictions. The main difference of the proposed ensemble model compared to conventional neural network models is that it is capable of getting a bootstrap confidence interval and probabilities of predictions. The performance of the proposed model is demonstrated on different time series. The obtained results show that the proposed ensemble model has a superior prediction performance in addition to having outputs that are more interpretable and applicable to probabilistic evaluations than conventional neural networks.
KW - bootstrap technique
KW - ensemble
KW - probabilistic prediction
KW - single multiplicative neuron model
KW - time series prediction
U2 - 10.1109/SSCI.2016.7849975
DO - 10.1109/SSCI.2016.7849975
M3 - Conference contribution/Paper
AN - SCOPUS:85016049289
T3 - 2016 IEEE Symposium Series on Computational Intelligence, SSCI 2016
BT - 2016 IEEE Symposium Series on Computational Intelligence, SSCI 2016
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2016 IEEE Symposium Series on Computational Intelligence, SSCI 2016
Y2 - 6 December 2016 through 9 December 2016
ER -