Visible to the public Time Series Forecasting by Evolving Deep Belief Network with Negative Correlation Search

TitleTime Series Forecasting by Evolving Deep Belief Network with Negative Correlation Search
Publication TypeConference Paper
Year of Publication2018
AuthorsLin, Y., Liu, H., Xie, G., Zhang, Y.
Conference Name2018 Chinese Automation Congress (CAC)
Date Publishednov
KeywordsBayes methods, Bayesian optimization, belief networks, Collaboration, composability, DBN, deep belief network, derivative-free optimizer-negative correlation search, Forecasting, forecasting theory, grid search, Human Behavior, hyperparameters optimization, learning (artificial intelligence), learning rates, Load modeling, Mathematical model, Metrics, Negative correlation search, neural nets, neural network models, Neural networks, optimisation, Optimization, policy-based governance, pubcrawl, random search, resilience, Resiliency, Scalability, search problems, time series, Time series analysis, time series datasets, time series forecasting, Training

The recently developed deep belief network (DBN) has been shown to be an effective methodology for solving time series forecasting problems. However, the performance of DBN is seriously depended on the reasonable setting of hyperparameters. At present, random search, grid search and Bayesian optimization are the most common methods of hyperparameters optimization. As an alternative, a state-of-the-art derivative-free optimizer-negative correlation search (NCS) is adopted in this paper to decide the sizes of DBN and learning rates during the training processes. A comparative analysis is performed between the proposed method and other popular techniques in the time series forecasting experiment based on two types of time series datasets. Experiment results statistically affirm the efficiency of the proposed model to obtain better prediction results compared with conventional neural network models.

Citation Keylin_time_2018