Comparative Study for Prediction of Neural Network models
Jinwoo Cho · 2018
In this study, ARIMA model and neural network models including MLP, CNN and RNN are compared according to their prediction power. Given assumption is that a neural network model would more suitable to the data that has regular patterns. To prove the assumption, the study set two simulations for stochastic process and deterministic process, which can represent randomness and visible pattern respectively. Finally, solution to develop the neural network model for time series is suggested regarding the result of simulations.