476 Pert tt Problems Principles in Practice 1, Suppose the monthly sales for «particular product for the past 20 months have been as follows: Moh 192 3 4 $ 6 7 8 9 © Saks 2am 0 5 6D Moh Mo 18 MIS 16 1 8 19 20 Saks 5055 4 48 55 47 615853 (2) Use a five-period moving average to compute forecasts of sales for months 6 to 20 and a seven period moving average to compute forecests for months 8 to 20, Which fs the data better for months 8 1020? Expl, (6) Use an exponential smoothing approach with smoothing constant a = 0.2 to forecast sales for months 2 20. Change 100.1. Does this make the fit better or worse? Explain, (©) Using exponential smoothing, find the value of e that minimizes the mean squared devistion (MSD) over months 2to 20. Find the value ofa that minimizes BIAS, Ace they th same? Explain, (@) Use an exponential smoothing with linear trend and smoothing constants = 0.4 and = 0.210 predict output for months 2to 20. Does this ft beter or worse then your answers to (0)? Explain, 2, The following data give closing values ofthe Dow Yones Industrial Average forthe 30 weeks, ‘months, and yeas prior to August 1, 1989. (@) Use exponential smoothing witha linear tend and smoothing coefficients of 0.1 on each set of data to generat forecast fo the Dow Jones Industrial ‘Average on August 1, 2000. Which data set do you think yields the best forecast? ‘What weight does a I-year-old data point get when we use smoothing constant a= 0.1 ‘on the weekly data? On the monthly data? On the annual data? What smoothing constant forthe monthly model that pves the same weight to I-year-old date is given by the ‘asnual model witha = 0.1? (©) Doss using the adjusted smoothing constant computed in par (b) (fo and f) in the monthly model make it predict more accurately the closing price for August 1, 2000? IF ot, why rot? (2) How much value do you think time series models have for forecasting stock prices? What features ofthe stock market make it difficult to predict, panticuaey inthe short term? ®) ‘Weekly Data | Monthly Data | Annual Data DateClose | Date Close a9 anpy 63777 | s59 ip ans7 65835 | smo 7646 ed anon SAM 898.1 insp9 sist siz 9637 259) sist BAN _8T6 2199 | Ta sim 6786 211589 wig Bans 353 2mp9 sn suns 9737 an1p9 soupy tana] arr s615 5899 uingy 78231] 978 9768 continued)Chapier 13 A QuoraSeaing Model a7 ‘Westy Data | Monthly Data | Annual Data Date Clove | Date Close | Date Close anse 99036] ransy 79083 | sano 8876 span 9g22| pe 79065 | amo 9325 sro oss | 2nws 387 | smi Ras ang soy7ss| snes 87098 | siz 9013 sna 10859} aps sacs | sma 12162 ang 109897] snws — sgc00 | see 1226 472698 10,788.0 wins 13360 5/99 110316 Bie 13983 51099 109133 mA) 26630 s/17p9 _ 10093 wins 2087 spa 10389:7 wine 2713 5399 10,7998 wiso 26148 78919805 SIAL 3.0486 ‘i149 108556 wig e249 _oss6 sn nyse 11,1392 4 39134 31599111987 snes 46106 saps 112098} se 105897 | sine 56162 71999 09110) 699 109708 | snan Tema ress 196851] res 106851 | 81498 7.5394 an 10740| nag 198293 | aria9 108293, {3 Hamburger Heaven has hired a team of stadents from the local university to develop a forecasting tool for predicting weekly burger sales to assist in the purchasing of supplies. The sssstant manager, who bes taken « couple of college clases, has heard of exponential smoothing and euggests that the sudents ry using it He gives them the following data on sales for the pas 16 weeks. Wee ot o2) Sales 3500 3,700 3,400 3900 3.500 3.600 4200 Wek 9 “oo 6 Ssles 9300 8.900 9,100 9.200 9.300 9,900 9,400 9,100, () What happens if exponential smoothing (ith no trend) is applied to these dat conventional manner? Use a smoothing constant c= 03. () Does itimprove the forecat if we use exponential smoothing wit a finear tend ead smoothing constants a = 6 = 0.37 (e) Suggest a modification of exponential smoothing that might make more sense for this situation 4, Select-1- Mode! offers computer-generated photos of people posing with famous supermodels ‘You simply send ina photo of yourself, and the company sends back a photo of you skiing, ot boating, ot night clubbing, or whatever, with a model, Of course, Selec-a-Model must pay the supermodels fr the use oftheir images. To anticipate cashflows, the company wants to ot up a forecasting system to predict sales. The following table gives monty demand forthe 3 past 2 years for three of the top-selling models.478 art Ill Prineiplesin Practice Month Model Model? Model 1 SS 2 3 2 3 (80 8 4% 86 55 ne) 2s 6 6 7 m6 ° 3 BO 8 9 97 Be © 2G " 2 lal ro m8 0 bm a 2 8 5% 0 7 96 6 8 7 1 6 7 a wo 36 8 10 “ 2» 10 2% 2 a n 2 10102 B nt 103, ™ mM (a) Plot the demand data for al three models and suggest forecasting mode! that might be suited to ech {b) Find suitable constants for model 1, How good a predictor isthe resulting model? {@) Find suitable constants for model 2. How good a predictor iste resuking model? (@) Find suitable eonstans for mode! 3. Tow good a predictor isthe resulting model? 55, Can-Do Canoe sels lightweight portable canoes. Quarterly demand forts most popular product family over the past 3 years has been as follows: _——————— Year 1996 997 1998 peseree aPC eee eye eee Pomme | 25 120 40 @ | 30 140 60 8} 35 IS SS (4) Use an exponential smoothing model with smoothing constant « = 0.2 10 develop a forecast for these data. How does itt? What isthe resulting MSD? () Use an exponential smoothing with linear trend mode! with smoothing constants a b.2 to develop a forecest for these data. How does itt? What is the resulting Msp? @) Use the Winters method with smoothing constanss « = f= y = 0.2 to develop @ orecas for these data, How does itt? What isthe resulting MSD? (6) Find smoothing constants that minimize MSD over the second two years of dats, How des the resulting forecast fit the data in the third year?Chapter 13 A QuotaSeing Modet a9 (@) Find smoothing constants that minimize MSD over the third year of data. How much ‘etter does the model fit the data in the third year than that of par (3)? Which model, (f) ‘9¢(@), do you think is likely to better predict demand in year 4” 6. Suppose a plant produces 50 customized high-performance bicycles per day and maintains oa average 10 days’ worth of WIP inthe system {Whats the average cycle time (time fiom when an order is eleased tothe plant uni the bicycle i complete, recy to ship)? () When would the conveyor model predict thet the 400th bicycle wil be competed? (© Suppose we currently have orders for 1,000 bieyces (i.e including the orders forthe 500 bicycles tht bave already been released tothe plant) and a customer is inguting about when we could deliver an order af 50 bicycles. Use the conveyor model to predict when this new order willbe completed. If we have flexibility conceming the due date we ‘quote to the customer, should we quote a date calculated earlier, ltr, oF a the same time as that computed using the conveyor model? Why? 7, Mareo, the manager ofa contractors supply store is concemed sbout predicting demand for the DeWally 519 hammer dil inorder to help plan for purchasing. He has brought in a team ‘of MBAs, who have suggested using a moving-sverage or exponential smoothing method. However, Marco is not sure this is the right approach because, as he points out, sales ofthe “sil ate affected by price. Since the store periodically runs promotions during which the price is reduced, he thinks that price should be accounted fo in the forecasting model. The following are price ad sles data forthe pas 20 weeks. Week rie Sales 195 mw 2 9924 m5 cee 26 ne) 198 12 1 w 6 30 38 9 7 38 rc) nim 4s B40 be 19 (2) Propose an altematve oa time series model for forecasting demand forthe DeWally 519. (b) Use your method forthe first weeks of dita to predict sales in week m + for n= 15,..., 19. How well does it work? (©) What does your model predic sles will be in week 21 if the pi i797 is $1992 the’ price is