Predictive Analysis is growing bigtime in the BI-Business Intelligence world. BI can be divided into three major tasks:
Scorecarding – is all about telling us where we are at now – how are we doing now and how is our competition doing. This is the bulk of BI with Reportwriting, OLAP & statistical analysis, and State of the Shop Portals being the bulk of applications in this arena.
Forecasting – is the real heart of Predictive Analysis. These are all the efforts applied to create causal models and predict future trends for an organization. EWS-Early Warning Systems for earthquakes, storms, stock market tumbles or take-offs are part of the forecasting business. But the bulk of work is forecasting future trends based on historical trends in the past (primarily univariate methods) or causal models which explain things in classic multivariate equation fashion – ProductSales = f(ConsumerIncome, CompetitorsSales, PopulationTrends,…).
Optimizing – decision making is the most complex BI activity as it derives much of its driving information from the previous Scorecarding and forecasting activities. There is a whole decision making and mathematical optimization theory with several hotly competing methods to master in this realm.
Forecasting and Predictive Analysis are important because a)they are so pervasive (everybody needs a good forecast)and b)with the storage, computing power, and data warehousing delivering low-cost enablers – the necessary conditions for doing forecasting are largely in place for many organizations from small to large. The software and methodologies are rapidly maturing in such fields as automated forecasting, data analysis, EWS-Early warning Systems, and multivariate causal modeling. Here is a list of resources:
Autobox – rugged software for automated and semi-automated univariate forecasting
SPSS – has gathered it stats, datamining, and times series software into Predictive Analytics
Forecast Pro – well designed univariate and multivariate forecasting software
Eviews – robust multivariate causal modeling toolset
Mikan – has interesting EWS forecasting software toolkit
SAS – has a broad array of statistical, automated forecasting, and multivariate analysis tools
Books and Link Resources
Forecasting Methods and Applications– One of the original surveys of forecasting methods for business
Time Series Models – for business forcasting and econometrics an excellent intro
The Analysis of Time Series – Chatfields books are consistently top rated, see why here
Data Mining – concepts and techniques with useful hands-on exercises
Business Modeling and Data Mining – looks at broader issues in buiness data mining
Modern Applied Statistics – broad overview of techniques as used with S, S-Plus and R stat packs
Modler – open source econometric modeling tool
Wharton Schools Principles of Forecasting – a book site that took on a life on its own on forecasting
Another Book that became a Great Website on Forecasting
Forecasting Software Survey – theOR/MS societys prestigious review of software
There are numerous threads in whats happening in forecasting because the enablers have just become available and that has stimulated a whole new layer of development on top of the classic techniques that were perfected in the 1970s thru to mid 1990s. Welcome to the new Predictive Analysis.