and forecasting methods. The novelty of the book is on integrating both methodologies and on the application of data mining for forecasting.
Software manual of SAS products – This is not an introductory manual of the SAS software products used in the application of data mining for forecasting. It is assumed that the interested reader has some basic knowledge on the specific SAS software used herein: Base SAS, SAS Enterprise Guide, SAS Enterprise Miner, and SAS Forecast Server.
Features of the Book
The key features that differentiate this book from other titles on data mining and forecasting are:
1 Integrating data mining and forecasting – One of the main messages in the book is that a critical factor for improving forecasting is using data mining methods. The synergetic benefits of both approaches are mostly in the area of variable reduction and variable selection for building multivariate forecasting models.
2 A broader view of industrial forecasting – Another important topic of the book is the proposed broadening of the forecasting approaches by using nonlinear predictions in addition to the existing time series methods. This allows handling cases with short time series and extraordinary business or process conditions.
3 Emphasis on practical applications – The third key feature of the book is the predominant practical view of all discussed topics. The examples given are from real industrial applications and the reader has the opportunity to “learn from the kitchen” regarding how data mining for forecasting works in an industrial setting.
Acknowledgments
The authors would like to thank Jan Baumgras and Terry Woodfield whose constructive comments substantially improved the final manuscript. The authors also highly appreciate the comments and clarifications of our technical reviewers Lorne Rothman, Abhijit Kulkarni, Sean Cai, Sara Vidal, and Udo Sglavo.
The staff of SAS Press has been most helpful, especially George McDaniel who successfully managed the project and responded to our requests. We gratefully acknowledge the contributions of our copyeditor Brad Kellam, production specialist Candy Farrell, designer Jennifer Dilley, and marketing specialists Aimee Rodriguez and Shelly Goodin.
Chapter 1: Why Industry Needs Data Mining For Forecasting
1.2 Forecasting Capabilities as a Competitive Advantage
1.3 The Explosion of Available Time Series Data
1.4 Some Background on Forecasting
1.5 The Limitations of Classical Univariate Forecasting
1.6 What is a Time Series Database?
1.7 What is Data Mining for Forecasting?
1.8 Advantages of Integrating Data Mining and Forecasting
1.1 Overview
In today's economic environment there is ample opportunity to leverage the numerous sources of time series data that are readily available to the savvy decision maker. This time series data can be used for business gain if the data is converted first to information and then to knowledge—knowing what to make when for whom, knowing when resource costs (raw material, logistics, labor, and so on) are changing or what the drivers of demand are and when they will be changing. All this knowledge leads to advantages to the bottom line for the decision maker when times series trends are captured in an appropriate mathematical form. The question becomes how and when to do so. Data mining processes, methods and technology oriented to transactional type data (data that does not have a time series framework) have grown immensely in the last quarter century. Many of the references listed in the bibliography (Fayyad et al. 1996, Cabena et al. 1998, Berry 2000, Pyle 2003, Duling and Thompson 2005, Rey and Kalos 2005, Kurgan and Musilek 2006, Han et al. 2012) speak to the many methods and processes aimed at building prediction models on data that does not have a time series framework. There is significant value in the interdisciplinary notion of data mining for forecasting when used to solve time series problems. The intention of this book is to describe how to get the most value out of the host of available time series data by using data mining techniques specifically oriented to data collected over time. Previous authors have written about various aspects of data mining for time series, but not in a holistic framework: Antunes, Oliveira (2006), Laxman, Sastry (2006), Mitsa (2010), Duling, Lee (2008), and Lee, Schubert (2011).
In this introductory chapter, we help build the case for using data mining for forecasting and using forecasting as a competitive advantage. We cover the explosion of available economic time series data, the basic background on forecasting, and the limitations of classical univariate forecasting (from a business perspective). We also define what a time series database is and what data mining for forecasting is all about, and lastly describe what the advantages of integrating data mining and forecasting actually are.
1.2 Forecasting Capabilities as a Competitive Advantage
Information Technology (IT) Systems for collecting and managing transactional data, such as SAP and others, have opened the door for businesses to understand their detailed historical transaction data for revenue, volume, price, costs and often times even the whole product income statement. Twenty-five years ago IT managers worried about storage limitations and thus would design “out of the system” any useful historical detail for forecasting purposes. With the decline of the cost of storage in recent years, architectural designs have in fact included saving various prorated levels of detail over time so that companies can fully take advantage of this wealth of information. IT infrastructures were initially put in place simply to manage the transactions. Today, these architectures should also accommodate leveraging this history for business gain by looking at it from an advanced analytics view point. Various authors have discussed this framework in detail (Chattratichat et al. 1999, Mundy et al. 2008, Pletcher et al. 2005, Duling et al. 2008).
Large corporations generally have many internal processes and functions that support businesses—all of which can leverage quality forecasts for business gain. This is beyond the typical supply chain need for having the right product at the right time for the right customer in the right amount. Some companies have moved to a lean pull replenishment framework in their supply chains. This lean approach does not preclude the use of high-quality forecasting processes, methods, and technology.
In addition to those who analyze the supply chain, many other organizations in a corporation can use high-quality forecasts. Finance groups generally control the planning process for corporations and deliver the numbers that the company plans against and reports to Wall Street. Strategy groups are always in need for medium- to long-range forecasts for strategic planning. Executive sales and operations planning (ESOP) demand medium-range forecasts for resource and asset planning. Marketing and sales organizations always need short- to medium-range forecasts for planning purposes. New business
