Online dynamic ensemble deep random vector functional link neural network for forecasting

• DOI: 10.1016/j.neunet.2023.06.042
• Publication Date: 2023-07-07T01:32:38Z
• Authors: Ruobin Gao; Ruilin Li; Minghui Hu; P.N. Suganthan; Kum Fai Yuen
• Keywords: Link (geometry)

This paper proposes a three-stage online deep learning model for time series based on the ensemble random vector functional link (edRVFL). The edRVFL stacks multiple randomized layers to enhance single-layer RVFL's representation ability. Each hidden layer's is utilized training an output layer, and of all forms edRVFL's output. However, original not designed learning, nature features harmful extracting meaningful temporal features. In order address limitations extend mode, this dynamic consisting three components, decomposition, training, ensemble. First, decomposition as feature engineering block edRVFL. Then, algorithm learn Finally, method, which can measure change in distribution, proposed aggregating layers' outputs. evaluates compares with state-of-the-art methods sixteen series.