Artificial Neural Network and Monte Carlo Simulation in a Hybrid Method for Time Series Forecasting with Generation of L-Scenarios

Abstract

Sometimes, there are time series segment, it is necessary to reconstruct information from the past, predict information for the future, in this paper a hybrid approach between Artificial Neural Network (ANN), Monte Carlo simulation (MCS) for the reconstruction (and / or prediction) of time series with the generation of L-scenarios is proposed, in order to evaluate results from hybrid method, the Chi-square test, analysis of variance (ANOVA), functions of autocorrelation were used, additionally, the forecasting ANN is compared with ARMAX model prediction, results show that the proposed method could reconstruct the past, could predict the future from known time series segment, so that each prediction in a whole period selected generates a scenario, the L-scenarios have high sameness statistical from original information. In the hybrid method, first, artificial neural network is trained with known information, second the statistics for the MCS are estimated, then L-scenarios were generated by MCS in the selected period, these information will serve such as inputs for ANN trained, finally these outputs ANN will be the whole time series within in the chosen period, which it want to be analysed.