Statistical Forecasting: Principles and Practice (TRA57)
Monday, October 21, 2013 at 2:00 PM - Wednesday, December 4, 2013 at 3:00 PM (PDT)
San Francisco, California
London, United Kingdom
This class is scheduled from October 21st through Dec 4th,
Monday and Wednesday
From 2:00pm -3:00pm PST.
We’ll take a break during Thanksgiving week.
(No class on 11/25th or 11/27th)
Instructor will be Rob Hyndman (http://robjhyndman.com/)
Statistical Forecasting: Principles and Practice
Forecasting is required in many situations: deciding whether to build another power generation plant in the next five years requires forecasts of future demand; scheduling staff in a call centre next week requires forecasts of call volume; stocking an inventory requires forecasts of stock requirements. Forecasts can be required several years in advance (for the case of capital investments), or only a few minutes beforehand (for telecommunication routing). Whatever the circumstances or time horizons involved, forecasting is an important aid in effective and efficient planning. In this workshop, we will explore methods and models for statistical forecasting of time series data.
Topics to be covered include seasonality and trends, exponential smoothing, ARIMA modelling, dynamic regression and state space models, as well as forecast accuracy methods and forecast evaluation techniques such as cross-validation. Some recent developments in each of these areas will be explored.
The text for the course is Hyndman & Athanasopoulos (2013), available online at OTexts.com/fpp.
The course will use the R packages fpp (Hyndman 2011) and forecast (Hyndman 2012).
Approximately 12 hours online lectures, plus exercises.
Users with some knowledge of R, and basic familiarity with statistics up to multiple regression, who would like to apply univariate time series forecasting methods in a business context.
• Familiarity with the basics of the R language and some prior hands-on experience.
• Understanding of multiple linear regression.
• Access to a computer running R or Revolution R Enterprise
Introduction to forecasting We will begin with some case studies frommy consulting practice. These are the sorts of problems we will be able to solve by the end of the course!
The forecaster’s toolbox Every forecaster needs a toolbox with a variety of tools to handle different types of problems. We will introduce a variety of graphical tools and some numerical techniques, and practice using them on simple forecasting problems in R.
Seasonality and trends Time series data frequently contain patterns that need to be captured by
forecasting models. We will discuss the differences between seasonality, cycles and trends, and
how to identify them.
Exponential smoothing Amongst the most popular, simple and accurate forecasting methods is
exponential smoothing and its extensions. We will look at its classical form, as well as the
modern formulation of exponential smoothing methods using state space models.
Residual diagnostics To ensure that a forecasting model has captured all of the available information, residual diagnostics are used. We will introduce the main diagnostic tools and how to apply them to any forecasting model.
Stationarity, transformations and differencing In preparation for ARIMA modelling, we will look
at some tools for transforming time series data to achieve stationarity.
ARIMA models The morning will be spent exploring the ARIMA modelling framework and applying
it to a range of different types of time series.
Time series cross-validation Cross-validation is a useful tool for measuring the predictive ability of
a model. We will look at how to apply it in a time series setting.
Dynamic regression Often, there will be additional information available to a forecaster that should be included in the model. One of the best ways of handling this information is via a dynamic
regression model. We will finish the workshop looking at these models and how to use them.
Hyndman, R. J. (2011), ‘fpp package for R’.
Hyndman, R. J. (2012), forecast: Forecasting functions for time series.
Hyndman, R. J. & Athanasopoulos, G. (2013), Forecasting: principles and practice, OTexts, Melbourne, Australia.
We have the right to cancel the event for any reason at any time. Revolution Analytics will refund all monies paid for ticket sales in full in the event of a cancellation. We are not responsible for any travel related expenses incurred by attendees for this event. This includes but not limited to transportation, hotel accommodations or any other travel related expenses secured by the attendee, due to a cancellation on our part.
- 30 or more days from the event date: Full refund less 10%
- 16-29 days from the event date: 50% refund
- 15 or less days from the event date: No refund
- all related transaction fees PayPal and Eventbrite are not refundable
- discount offers cannot be combined
- A student ID Number is not a proof of full time university enrollment to get the student’s discount. Proof of enrollment in 9 units or more on a current academic registration document will be required to receive the student's discount.