Introduction to Forecasting

Data Science Ballroom AB
Average rating: ****.
(4.25, 16 ratings)

Proper forecasting is crucial to Facebook’s business operations. We rely on reliable forecasts in deciding how many people to hire and where, in making multi-million dollar investment decisions, and determining the allocation of servers and bandwidth.

I have led Facebook’s demand and advertising forecasting effort for the past year and previous to Facebook developed several forecasting models including a model of future international student enrollments, named `ISAFM`, for ICG consulting group: http://www.illuminateconsultinggroup.biz/isafm/.

I have found in my experience that forecasting is a less accessible topic than most other statistical inference techniques. It is often done in an ad-hoc manner and best practices are commonly violated. Too often big decisions are made without any forecasting at all.

Forecasting has also been given the reputations as being more of an art than a science and as only being appropriately applied by domain experts. In reality, every data analyst should have forecasting in their toolkit to know what the data tells us about the probabilities of future scenarios. Domain expertise is crucial for appropriately interpreting forecasting models, but this should not present a barrier to entry for forecasting newbies. Instead, forecasters should gather all the information the data contains and seek out as much domain knowledge as possible to appropriately interpret the output.

In this session, I will cover the basic of forecasting to make forecasting accessible to every data analyst. I plan to cover:

Basic techniques
Model Selection
Model Validation
Difference between forecasting and prediction
Best Practices
Tips
Simple examples in R

I may also cover forecasting lessons learned at Facebook and Facebook’s forecasting process.

Photo of Michael Bailey

Michael Bailey

Facebook

Michael Bailey is an Economist and Researcher at Facebook and leads advertising demand forecasting. He joins a small group of economists specializing in Big Data economics and developing distributed economics algorithms. He graduated from Stanford with a PhD in Economics where he studied applied econometrics and internet economics. While a graduate student, he developing forecasting models for businesses including the International Student Forecasting Analytics Model (ISAFM).

Comments on this page are now closed.

Comments

Paul Bunkham
04/14/2013 11:28pm PDT

Thanks that would be great. I’ve sent an email a few days ago, but thought I’d put a comment on here in case it falls in to your junk folder.

Picture of Michael Bailey
Michael Bailey
04/10/2013 4:08pm PDT

@Paul, shoot me an email (mikethechampion at gm ail.com) and I can send you the slides with links directly.

Paul Bunkham
04/10/2013 12:41am PDT

Interesting session. I’ve just tried to look at the references in your presentation and I think the links have been removed. Any chance you could post one with the URLs?

Sponsors

Sponsorship Opportunities

For information on exhibition and sponsorship opportunities at the conference, contact Susan Stewart at sstewart@oreilly.com

Media Partner Opportunities

For information on trade opportunities with O'Reilly conferences contact Kathy Yu at mediapartners
@oreilly.com

Press and Media

For media-related inquiries, contact Maureen Jennings at maureen@oreilly.com

Contact Us

View a complete list of Strata contacts