Open AccessA Novel Hybrid Mgstar-Rnn Model for Forecasting Spatio-Temporal Data
Author(s) -
V. O. N. Laily,
Suhartono Suhartono,
Elly Pusporani,
Raden Mohamad Atok
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1752/1/012011
Subject(s) - autoregressive model , computer science , component (thermodynamics) , time series , dimension (graph theory) , series (stratigraphy) , recurrent neural network , nonlinear system , artificial intelligence , machine learning , mathematics , econometrics , artificial neural network , pure mathematics , paleontology , physics , quantum mechanics , biology , thermodynamics
Time series data have a time dimension and space dimension called Spatio-temporal data. GSTAR is one of the models that can be used to analyze spatio-temporal data. One of development for this model is the Multivariate Generalized Spaced-Time Autoregressive (MGSTAR). The MGSTAR model has limitations of not being able to model a nonlinear time series, this problem can be overcome by applying a hybrid model on MGSTAR. This research aims to propose a hybrid MGSTAR-RNN model, where the MGSTAR model as a linear component and RNN as a nonlinear component. This research focused on a simulation study to evaluate the performance of MGSTAR and MGSTAR-RNN model. Several scenarios be experimented, i.e simulation in data with different variable component and location. The result show that hybrid MGSTAR-RNN model is better than individual MGSTAR model. In general, it is in line with the latest results of the 2018 M4 forecasting competition show that combined models or hybrid models tend to provide more accurate forecast performance than forecast results with individual models.