### Abstract

Time series forecasting is usually done in a deterministic sense, such as in autoregressive moving average models, where a future state is predicted as a linear combination of past events. However, by formulating the problem in a probabilistic sense, soft predictions are obtained from a given probability mass function. This paper uses a deep neural network for probabilistic forecasting of time series by minimizing the cross entropy of the probability of future symbols from a given state. The advantage of this type of model is that it makes probabilistic inferences from the ground up, and without any restrictive assumptions (e.g., second order statistics). The efficacy of the proposed model is tested by forecasting the emergence of combustion instabilities, defined to be the root mean square of the pressure signal inside a laboratory-scale combustor system. The proposed algorithm has been compared with the autoregressive moving average (ARMA) model, which acts as a baseline for many time-series forecasting tasks, and the proposed model is shown to significantly outperform the ARMA model in this task.

Original language | English (US) |
---|---|

Title of host publication | 2017 American Control Conference, ACC 2017 |

Publisher | Institute of Electrical and Electronics Engineers Inc. |

Pages | 3147-3152 |

Number of pages | 6 |

ISBN (Electronic) | 9781509059928 |

DOIs | |

State | Published - Jun 29 2017 |

Event | 2017 American Control Conference, ACC 2017 - Seattle, United States Duration: May 24 2017 → May 26 2017 |

### Other

Other | 2017 American Control Conference, ACC 2017 |
---|---|

Country | United States |

City | Seattle |

Period | 5/24/17 → 5/26/17 |

### Fingerprint

### All Science Journal Classification (ASJC) codes

- Electrical and Electronic Engineering

### Cite this

*2017 American Control Conference, ACC 2017*(pp. 3147-3152). [7963431] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.23919/ACC.2017.7963431

}

*2017 American Control Conference, ACC 2017.*, 7963431, Institute of Electrical and Electronics Engineers Inc., pp. 3147-3152, 2017 American Control Conference, ACC 2017, Seattle, United States, 5/24/17. https://doi.org/10.23919/ACC.2017.7963431

**Probabilistic forecasting of symbol sequences with deep neural networks.** / Hauser, Michael; Fu, Yiwei; Li, Yue; Phoha, Shashi; Ray, Asok.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Probabilistic forecasting of symbol sequences with deep neural networks

AU - Hauser, Michael

AU - Fu, Yiwei

AU - Li, Yue

AU - Phoha, Shashi

AU - Ray, Asok

PY - 2017/6/29

Y1 - 2017/6/29

N2 - Time series forecasting is usually done in a deterministic sense, such as in autoregressive moving average models, where a future state is predicted as a linear combination of past events. However, by formulating the problem in a probabilistic sense, soft predictions are obtained from a given probability mass function. This paper uses a deep neural network for probabilistic forecasting of time series by minimizing the cross entropy of the probability of future symbols from a given state. The advantage of this type of model is that it makes probabilistic inferences from the ground up, and without any restrictive assumptions (e.g., second order statistics). The efficacy of the proposed model is tested by forecasting the emergence of combustion instabilities, defined to be the root mean square of the pressure signal inside a laboratory-scale combustor system. The proposed algorithm has been compared with the autoregressive moving average (ARMA) model, which acts as a baseline for many time-series forecasting tasks, and the proposed model is shown to significantly outperform the ARMA model in this task.

AB - Time series forecasting is usually done in a deterministic sense, such as in autoregressive moving average models, where a future state is predicted as a linear combination of past events. However, by formulating the problem in a probabilistic sense, soft predictions are obtained from a given probability mass function. This paper uses a deep neural network for probabilistic forecasting of time series by minimizing the cross entropy of the probability of future symbols from a given state. The advantage of this type of model is that it makes probabilistic inferences from the ground up, and without any restrictive assumptions (e.g., second order statistics). The efficacy of the proposed model is tested by forecasting the emergence of combustion instabilities, defined to be the root mean square of the pressure signal inside a laboratory-scale combustor system. The proposed algorithm has been compared with the autoregressive moving average (ARMA) model, which acts as a baseline for many time-series forecasting tasks, and the proposed model is shown to significantly outperform the ARMA model in this task.

UR - http://www.scopus.com/inward/record.url?scp=85027035902&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85027035902&partnerID=8YFLogxK

U2 - 10.23919/ACC.2017.7963431

DO - 10.23919/ACC.2017.7963431

M3 - Conference contribution

SP - 3147

EP - 3152

BT - 2017 American Control Conference, ACC 2017

PB - Institute of Electrical and Electronics Engineers Inc.

ER -