The performance of a Recurrent HONN for temperature time series prediction

An ARMA Type Pi-Sigma Artificial Neural Network for Nonlinear Time Series Forecasting

“…Let (c 1i , c 1 f ), (c 2i , c 2 f ), and (w 1 , w 2 ) be the intervals which includes possible values for c 1 , c 2 and w, respectively. At each iteration, these parameters are calculated by using the formulas given in (7), (8) and (9).…”

“…In the literature, there are several recurrent architectures for PS-ANN. [8,16,17] proposed Jordan Type recurrent PS-ANN. In these methods, the output of recurrent PS-ANN is linked to the input layer as one step lagged and shown as a new input.…”

“…The updated values of cognitive coefficient c 1 , social coefficient c 2 , and inertia parameter w are calculated using the formulas given in (7), (8), and (9).…”

“…While Egrioglu et al [2] proposed recurrent multiplicative neuron model artificial neural network (RMNM-ANN), Gundogdu et al [3] PS-ANN proposed by Shin and Ghosh [7] is also a kind of higher order neural network. Ghosh and Shin [7] argued that PS-ANN requires less memory (weights and nodes), and at least two orders of magnitude less number of computations when compared to MLP-ANN for similar performance level, and over a broad class of problems [8]. [9,10,11] used PS-ANN for time series forecasting problem.…”

“…Moreover, there are some recurrent Pi-Sigma neural networks (R-PS-ANN) in the literature. Ghazali et al [8] introduced an R-PS-ANN that the output of NN is connected to input layer as one-step-lagged and forms a new input. [16] and Nayak et al [17] presented Jordan type R-PS-ANN.…”