Improving multi-step time series prediction with recurrent neural modelling

Multi-step prediction is a difficult task ¡hat has been attracted increasing tbe inieres! in recen! years. It tries to achieve predictions several sleps ahead ¡nto the tuture starting from information al time k. This paper is facllsed on the dcvelopment oí nonlinear neural models with tbe purpose oí building long-teTm Uf multi-step time series prediction schemes. In these context, the mos! popular neural models are bascd on the traditional feedforward neural network. However, ¡hese kind oí models may presen! SOrne problems when a long-term prediction problem is formulated. In tbis paper, a nenTal model based un a partially recurrent neural network is proposed as an altemative. Por fue new model, a learning phase with the purpose of long-term prediction is imposed, which allows to obtain bettcr predictions of time series in the future. This recurrent neural model has beeo applied to the logislic time series with the aim to predict the dynamic behaviour Di the series in the future. Mode1s based 00 feedforward nemal networks have beeo also llsed and compared against lhe proposed model.