A Neural Networks Approach for DORIS Time Series Prediction
Event: 2018 IDS Workshop
Session: SESSION I: DORIS network and constellation: status and evolution
Presentation type: Type Poster
Contribution: not provided
Generally, permanent station time series also include various types of signals, as both real and apparent causes (such as miss-modeled errors, effects of observational environments, random noise or any other effects produced by analysis software and settings of a prior stochastic models). Data analysis to the station time series aims to extract useful signals, such as crustal deformation, seasonal variations of station dynamics etc. During the past few years numerous models for analyzing and forecasting of time series have been developed by researchers. The study investigates a possibility to utilize artificial neural networks (ANN) in DORIS time series seasonal component analysis. Multilayer perceptron model is proposed for time series forecasting. The series of weekly SINEX solutions grgwd40 and ign17wd05, provided by GRG and IGN Analysis Centers respectively were used for analysis.