Time series forecasting based on parallel neural network

Abstract

In this paper we show a Parallel Neural Network (Cross-over Prediction Model) for time series forecasting implemented in PVM (”Parallel Virtual Machine”) and MPI (”Message Passing Interface”), in order to reduce computational time. Parallelization is achieved twofold: (a) updating autoregressive parameters using a genetic algorithm (GA) and (b) evaluating the overall prediction function via a parallel neural network. We implement the GA in two popular architectures of parallel processors (i.e hypercube and 2D-mesh) and discuss their time efficiency.