Several novel evaluation measures for rank-based ensemble pruning with applications to time series prediction

A hybrid model for high-frequency stock market forecasting

“…Artificial neural networks (ANNs) can be viewed as an alternative promising approach, regarding statistical models, for time series forecasting (Aznarte, Alcala-Fdez, Arauzo-Azofra, & Benítez, 2012;de Oliveira, Zarate, & Nobre, 2011;Cheng & Wei, 2014;Feng & Chou, 2011;Khashei & Bijari, 2014;Kourentzes, Barrow, & Crone, 2014;Kristjanpoller, Fadic, & Minutolo, 2014;Ma, Dai, & Liu, 2015;Menezes & Barreto, 2013;Nassirtoussi, Aghabozorgi, Wah, & Ngo, 2014;Park, Kim, & Lee, 2014). However, it has a main limitation, which is the definition of the parameters set (topology, architecture, amount of hidden units within hidden layer and training algorithm are some of these parameters) .…”

Predicting long-term population dynamics with bagging and boosting of process-based models

“…This is a standard approach for 23 improving the predictive performance of models in machine learning Diet-24 terich (2000). An ensemble is a set of models, referred to as base models (Ma et al, 2015), they have so far been used to make 33 short-term predictions, i.e., predict only the values of the system variables 34 in immediate future and not the long-term system behavior. 35 The main motivation for the work presented in this paper is to improve 36 the accuracy of long-term predictions of process-based models.…”

Prediction of liquid ammonia yield using a novel deep learning‐based heterogeneous pruning ensemble model

“…Ma et al mentioned that the prediction error of time series data is “directional.” That means even if the prediction errors of models against real data are comparable in magnitude, the trend of their prediction curves may vary drastically. Therefore, guaranteeing both small prediction error of the base models and large direction diversity of each other enhances complementarity among the models in the ensemble.…”