Ensembles of Randomized Neural Networks for Pattern-based Time Series Forecasting

This is a video presentation for the paper "Ensembles of Randomized Neural Networks for Pattern-based Time Series Forecasting". The work was presented at the 28th International Conference on Neural Information Processing (ICONIP2021), December 8-12, 2021, Bali, Indonesia (iconip2021.apn....
The paper: arxiv.org/abs/...
Abstract:
In this work, we propose an ensemble forecasting approach based on randomized neural networks. Improved randomized learning streamlines the fitting abilities of individual learners by generating network parameters adjusted to the target function complexity. A pattern-based time series representation makes the ensemble model suitable for forecasting problems with multiple seasonal components. We develop and verify forecasting models with six strategies for controlling the diversity of ensemble members. Case studies performed on several real-world forecasting problems verified the superior performance of the proposed ensemble forecasting approach. It outperformed both statistical and machine learning models in terms of forecasting accuracy. The proposed approach has several advantages: fast and easy training, simple architecture, ease of implementation, high accuracy and the ability to deal with nonstationarity and multiple seasonality.
Keywords:
Ensemble forecasting, Pattern representation of time series, Randomized neural networks, Short-term load forecasting, Time series forecasting.