Grant: Discovery Grants Program - Individual
University of Waterloo
$16,000.00
Modelling and forecasting time series using autoregressive approximation
FY 2012-13
Summary
Forecasting the future is one of the most important activities in our life. Every day, inventory, production, scheduling of personnel and machinery, financial and marketing decisions are made which depend on short, medium and long-term forecasts. A good forecaster must be able to quickly identify a bridge between the available data and the most relevant modeling technique, provide the reliable inference about the future and to adapt the forecasting model to new changes in the world. In many practical situations the sample size of observations is not known a-priori and may indefinitely increase while a data analyst still needs to perform prediction tasks in real time. Such cases are typically referred to as online modelling and forecasting and are widely met in a variety of modern applications, e.g. prediction of stock or currency exchange returns, filtering electrocardiogram measurements of heart rate etc. Arrival of new observations often implies that the order of the currently utilized model should be refined and, thus, all model parameters need to be recalculated. Hence, computational costs eventually increase. This is especially important if the underlying model has a complicated structure.
The goal of the proposed research is to investigate theoretical and practical properties of approximating more complicated model structures by simpler autoregressive models, which parameters are estimated using the Regularized Least Squares method. Among the potential benefits of the proposed methodology are computational efficiency, parsimonius model structure, wide applicability to a variety of time series that are of particular practical importance nowadays, e.g. long memory processes. The proposed methodology has common threads with shrinkage and ridge regressions and involves such statistical areas as analysis of long range dependent and periodic time series, optimal model selection and elements of information theory, Bayesian inference and generally theory of estimation and inference for stochastic processes. Potential applications include financial time series analysis, signal and image processing, weather and climate modelling.
Program
Discovery Grants Program - Individual
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
312350-2008
Name nom
Gel, Yulia
Department département
Statistics and Actuarial Science
Organization id
33
Institution établissement
University of Waterloo
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2012
Competition year année de concours
2008
Award amount
16000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
14
Committee name en
Statistical Sciences
Committee name fr
Statistique
Area of application code
405
Area of application group en
Environment
Area of application group fr
Environnement
Area of application en
Modelling and mathematical simulation of natural processes
Area of application fr
Modélisation et simulation mathématique des processus naturels
Research subject code
3002
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Time series analysis
Research subject fr
Analyse des séries chronologiques
Partie
5
Nb partie
5
Application title
Modelling and forecasting time series using autoregressive approximation
Keyword
NSERC
Application summary
Forecasting the future is one of the most important activities in our life. Every day, inventory, production, scheduling of personnel and machinery, financial and marketing decisions are made which depend on short, medium and long-term forecasts. A good forecaster must be able to quickly identify a bridge between the available data and the most relevant modeling technique, provide the reliable inference about the future and to adapt the forecasting model to new changes in the world. In many practical situations the sample size of observations is not known a-priori and may indefinitely increase while a data analyst still needs to perform prediction tasks in real time. Such cases are typically referred to as online modelling and forecasting and are widely met in a variety of modern applications, e.g. prediction of stock or currency exchange returns, filtering electrocardiogram measurements of heart rate etc. Arrival of new observations often implies that the order of the currently utilized model should be refined and, thus, all model parameters need to be recalculated. Hence, computational costs eventually increase. This is especially important if the underlying model has a complicated structure.
The goal of the proposed research is to investigate theoretical and practical properties of approximating more complicated model structures by simpler autoregressive models, which parameters are estimated using the Regularized Least Squares method. Among the potential benefits of the proposed methodology are computational efficiency, parsimonius model structure, wide applicability to a variety of time series that are of particular practical importance nowadays, e.g. long memory processes. The proposed methodology has common threads with shrinkage and ridge regressions and involves such statistical areas as analysis of long range dependent and periodic time series, optimal model selection and elements of information theory, Bayesian inference and generally theory of estimation and inference for stochastic processes. Potential applications include financial time series analysis, signal and image processing, weather and climate modelling.