Mackey-Glass noisy chaotic time series prediction by a swarm-optimized neural network

摘要

Inthisstudy,anartificialneuralnetwork(ANN)basedonparticleswarmoptimization(PSO)wasdevelopedforthetimeseriesprediction.ThehybridANN+PSOalgorithmwasappliedonMackey-Glassnoiselesschaotictimeseriesintheshort-termandlong-termprediction.Theperformancepredictionisevaluatedandcomparedwithsimilarworkintheliterature,particularlyforthelong-termforecast.Also,wepresentpropertiesofthedynamicalsystemviathestudyofchaoticbehaviourobtainedfromthetimeseriesprediction.Then,thisstandardhybridANN+PSOalgorithmwascomplementedwithaGaussianstochasticprocedure(calledstochastichybridANN+PSO)inordertoobtainanewestimatorofthepredictionsthatalsoalloweduscomputeuncertaintiesofpredictionsfornoisyMackey-Glasschaotictimeseries.Westudytheimpactofnoiseforthreecaseswithawhitenoiselevel(σN)contributionof0.01,0.05and0.1...