Conference/Workshop of interest for AIROYoungers
Dates
From: 21 Sep 2020
To: 25 Sep 2020
To: 25 Sep 2020
Location
Lancaster University
Country
UK
Description
Introduction
The week long course will focus on the principles of time-series forecasting, and recent developments in predictive analytics. Many examples will be used and there will be opportunities for participants to use models ‘hands-on’ and to experiment with real data. Links between forecasting, predictive analytics and decision making under uncertainty will be emphasised. The relevance of forecasting to non-stationary environments will be discussed, for example in inventory, simulation and optimisation models.
Location and Session Leaders
Sessions will be led by experts from the Lancaster Centre for Forecasting, calling on external experts if necessary. The overall co-ordinator will be Professor John Boylan.
Course Structure
The course will be subdivided into three sections, as follows:
•Time series modelling
•Econometric models
•Predictive analytics
This forms a natural sequence, moving from univariate models to classical multivariate models to modern methods which address both univariate and multivariate data.
The session on time series modelling will commence with the simplest methods, with opportunities for participants to experiment with parameter choices and to appreciate the scope for parameter optimisation methods to improve forecast accuracy. Model-based approaches will be covered, as well as criteria for choosing between forecasting models. A more general discussion will be held on error measurement and concerns about the reproducibility of forecasting research.
The session on econometric models will start with an introduction to the concepts involved, but will assume some knowledge of regression modelling. It will cover diagnostic analysis of regression models and model choice. This session will also cover some of the important technicalities of econometric time-series models, such as spurious regression, cointegration and modelling non-stationary data. It will conclude with a broader discussion of asymmetric loss functions and forecaster behaviour.
The final session on predictive analytics will cover some key topics for time series analysis. This will include data mining, with a focus on time series clustering for data exploration, and time series classification. It will also cover artificial intelligence for forecasting, focussing on Artificial Neural Networks and recent developments in ensemble methods for forecasting.
The week long course will focus on the principles of time-series forecasting, and recent developments in predictive analytics. Many examples will be used and there will be opportunities for participants to use models ‘hands-on’ and to experiment with real data. Links between forecasting, predictive analytics and decision making under uncertainty will be emphasised. The relevance of forecasting to non-stationary environments will be discussed, for example in inventory, simulation and optimisation models.
Location and Session Leaders
Sessions will be led by experts from the Lancaster Centre for Forecasting, calling on external experts if necessary. The overall co-ordinator will be Professor John Boylan.
Course Structure
The course will be subdivided into three sections, as follows:
•Time series modelling
•Econometric models
•Predictive analytics
This forms a natural sequence, moving from univariate models to classical multivariate models to modern methods which address both univariate and multivariate data.
The session on time series modelling will commence with the simplest methods, with opportunities for participants to experiment with parameter choices and to appreciate the scope for parameter optimisation methods to improve forecast accuracy. Model-based approaches will be covered, as well as criteria for choosing between forecasting models. A more general discussion will be held on error measurement and concerns about the reproducibility of forecasting research.
The session on econometric models will start with an introduction to the concepts involved, but will assume some knowledge of regression modelling. It will cover diagnostic analysis of regression models and model choice. This session will also cover some of the important technicalities of econometric time-series models, such as spurious regression, cointegration and modelling non-stationary data. It will conclude with a broader discussion of asymmetric loss functions and forecaster behaviour.
The final session on predictive analytics will cover some key topics for time series analysis. This will include data mining, with a focus on time series clustering for data exploration, and time series classification. It will also cover artificial intelligence for forecasting, focussing on Artificial Neural Networks and recent developments in ensemble methods for forecasting.