BUS 655/655/455 Forecasting & Predictive Analytics How to Predict the Future with Confidence Spring 2015 Wednesday 6:30-9:15, Room 208 Prof. Stephen P. Stuk, Ph.D. Office GBS 409 Mobile: 678-468-3452 use this one Stephen.Stuk@emory.edu DIA area administrator - Jalisa Norton: 404-727-8698 Books: Business Forecasting, 9/E,John E. Hanke & Dean Wichern :ISBN-10: 0132301202 ISBN-13: 9780132301206, Publisher: Prentice Hall, Copyright: 2009 http://www.pearsonhighered.com/educator/product/business-forecasting-9e/9780132301206.page This course is about a large variety of ways to analyze data and generate mathematical models to predict new or future values. Fortunately many of skills and models that we will be discussing can be applied to a plethora of scenarios. It is likely that a technique applied to predict the volume of sales next year will help predict who is likely to buy a new product. While we will be concentrating on process and techniques we will be discussing examples of structure. I hate the term data mining in that it implies looking for structure with statistics. I am more interested in confirming structural behavior with what we find in the data. This is not a minor point. To this end the class is VERY hands on, we are attempting to experience years of modeling in 3 months. To do that we must all PLAY with datasets and share what we find, both what worked and what didn t. Start thinking about what data or application you want to start looking at. If you have data from a project, another class, or work, that would be great. Each class session will be equally divided between presentation of new material and discussion of the application, by me and by you of ongoing applications. Grading: 40% - 4 projects (10% each), 10% - weekly quizzes 10% Class Participation (half sharing, half presentation), 20% weekly assignments (there will be an excel file to turn in electronically) 10% final presentation. 10% final quiz.
Software: Excel JMP ( free off the servers - see end of syllabus) Neural Network Software ( JMP, NeuralTools, others) BUS 655 Preliminary Class Schedule Session #1 (January 14) No need to have book in advance. Bring your computer; we will do lots of things together in class, hands on, as will happen most weeks. Introduction, what we are going to do, why we are going to do it. We need more than just models and numbers we need something to help us utilize the models and predictions to make better decisions. Time series and cross-sectional data. Decision Support Systems DSS. Optimization Metric Lots of examples Discussion of Data What do I need to do, what technique should I use, how do I know if it works? Review Regression as you know it. Transformations, Trend, Time dependent. Quality of fit. Residuals. Look at Excel to generalize model fit. Use of Solver, beginning of a DSS Excel, JMP Session #2 ( January 21) Reading: Look over Chapter 1-Review & 2-Statistics Review After class Look at Chapter 3-Exploring Data patterns Especially pages 81-84 & table on page 80 BUS655/655P/455 Spring 2015 - Draft Syllabus p. 2
Problems: chapter 2 page 48 #11 Chapter 3 page 94 case 3-1A (Murphy Brothers) read and calculate measures of fit Use a lag 12 period simple forecast, Introduction to Moving Averages and exponential smoothing. Session #3 (January 28) Read Chapter 4 Page 108 Naïve models Page 111 moving average Page 119 Exponential smoothing ( also see handout from class) Page 136 Formulas Problems: Page 138 9 a-h page 148 Chapter 4 case 4-4 Moving Average and exponential Smoothing Introduction of Moving averages, exponential smoothing Moving average: simple, length, centered Exponential models, simple, double(trend), seasonal ( winters) There will be a handout on these methods and Excel templates. Session #4 (February 4) Problems: Using the data from the Murphy case Apply, simple, holts, winters models. Seasonal models - regression? Examples of Moving Average and exponential smoothing Data Mining tools. BUS655/655P/455 Spring 2015 - Draft Syllabus p. 3
Session #5 (February 11) First Look at Synthetic Neural Networks an Introduction More examples of Moving Average and exponential smoothing Simple, Holts, Winters exponential smoothing continued, including multiplicative models. Session #6 (February 18) Autocorrelation methods Introduction, More Data Mining tools. Introduction of autocorrelation, cross-correlation, Box-Jenkins models, ARIMA models Session #7 (February 25) Examples and hands on time ******** 1 first model due ********** No class March 4, ************** Session #8 (March 11) This is a BBA session, MBA s in town are welcome Topics to be determined. Session #9 (March 18) Neural Networks visit II More techniques with ARIMA GARCH Bob Engle BUS655/655P/455 Spring 2015 - Draft Syllabus p. 4
Examples and analysis ******* 2 Second model due ********** Session #10 (March 25) Decision Support Systems DSS actually use the model. Advanced Network examples Recurrent networks, Kohonen networks Session #11 ( April 1) How to attack a new problem Data mining/exploration path Using Social Media Data Different versions of SNN s Example of clustering with/without SNN s ******** 3 third model due ********** Session #12 (April 8) Chaos theory and Fractiles Spectral Analysis, etc. Presentation of additional applications Session #13 (April 15) The future Where is it going. Emerging tools and techniques BUS655/655P/455 Spring 2015 - Draft Syllabus p. 5
Wrap up material Student presentations ******** 4 Final model should be done by now ********** Session #14 (April 22) Student Presentations Take home final Quiz ( TBD ) BUS655/655P/455 Spring 2015 - Draft Syllabus p. 6
Grading will be based on A Class participation 1) sharing in discussions 2) presenting your work B Four models 1) Nonlinear regression 2) Exponential smoothing 3) Neural Network. 4) Decision Support System C D E Final Presentation of your work Quizzes ( weekly and final take home) Simple assignments each week to keep up to speed. JMP: Jmp is statistical software created by the SAS Institute (the makers of SAS). It is available for free on the GBS network. To download it, get connected to the GBS network and go to the swmisc on 'gbsfile' drive (on my computer it is drive U). You will see folders for WINDOWS and Mac so navigate to the folder corresponding to your computer s operating system. The location of the JMP software for Windows users is: U:\WINDOWS\JMP\Windows Installation Files\JMP For Mac users it is U:\Mac\JMP\Macintosh Installation Files Windows uses should double click on the file setup.exe and follow the prompts. I am not familiar with the installation method for Mac software. If you have trouble installing the software, please contact the help desk. We will be using EXCEL, JMP and Q-net a great deal. The Reference Texts are: 1) Business Forecasting, Hanke & Reitsch. 2) Multivariate Data Analysis, Hair,Anderson,Tatham Black, Prentice Hall, 1998, ISBN 0-13-894858-5 BUS655/655P/455 Spring 2015 - Draft Syllabus p. 7
3) Basic Business Statistics, 7 th or 8 th edition by Berenson & Levine 4) Data Analysis & Decision Making with Microsoft Excelby Albright Winston and Zappe Duxbury Press SBN 0-534-26124-8 I will be providing extensive notes. Look/Ask about data sets of interest to you. List of URL s to be provided shortly. BUS655/655P/455 Spring 2015 - Draft Syllabus p. 8