深層学習と線形モデルを併用した時系列予測手法

カテゴリ: 部門大会

論文No: TC3-5

グループ名: 【C】平成27年電気学会電子・情報・システム部門大会講演論文集

発行日: 2015/08/27

タイトル(英語): A Novel Approach to Time Series Forecasting using Deep Learning and Linear Model

著者名: 平田 貴臣(山口大学),呉本 尭(山口大学),大林 正直(山口大学),間普真吾 (山口大学),小林 邦和(愛知県立大学)

著者名(英語): Takaomi Hirata(Yamaguchi University),Takashi Kuremoto(Yamaguchi University),Masanao Obayashi(Yamaguchi University),Shingo Mabu(Yamaguchi University),Kunikazu Kobayashi(Aichi Prefectural University)

キーワード: 時系列予測|自己回帰移動平均モデル|深層学習|ディープビリーフネット|制限付きボルツマンマシンカオス時系列|time series forecasting|ARIMA model|deep learning|deep belief net|restricted Boltzmann machinechaotic time series

要約(日本語): Since 1970s, linear models such as ARIMA have been popular for time series data analyze and prediction. Meanwhile, artificial neural networks (ANNs), which are nonlinear models, inspired by connectionism bio-informatics, have been showing their powerful abilities of function approximation, pattern recognition, dimensionality reduction, and so on. Recently, deep belief nets (DBNs) which use multiple Restricted Boltzmann machines (RBMs) and multi-layered perceptron (MLP) are proposed as time series predictors. In this study, a hybrid prediction method with DBN and ARIMA is proposed. The effectiveness of the novel method was confirmed by the experiments using chaotic time series data, and CATS benchmark data.

PDFファイルサイズ: 1,027 Kバイト