Industrial and Production Engineer BSC Eng. (SUST), MEng.(BUET) E-Mail Mamun_ipe@yahoo.com Mob: 01730739318 11/27/2013 1 Forecasting Lecture outline: Introduction Forecasting time horizons Types of forecasts Seven steps in the forecasting system Demand Patterns Forecasting approaches Qualitative approaches Quantitative approaches Measure of forecasting accuracy Example 11/27/2013 2 2 Forecasting time horizons Short rangeforecast: Thisforecast hasatimespanof upto1year but isgenerallylessthan3months. It is used for planning, job scheduling, workforce levels, job assignmentsandproductionlevels. Medium range forecast: A medium-range or intermediate, forecast generallyspansfrom3monthsto3years. It is used for sales planning, production planning and budgeting, cashbudgetingandanalyzingvariousoperatingplans. 11/27/2013 3 Introduction The art and science of predicting future events. Long range forecast: Generally 3 years or more in time span, long-range forecasts are used in planning for new products, capital expenditures, facility location or expansion and research and development 11/27/2013 4 3 Types of forecasts 1.Typesof Forecastinginbusiness: Organizations use the three major types of forecasts in future operationsplanning: 1.Economic forecasts address the business cycle by predicting inflation rates, money supplies, housing starts, and other planning indicators 2.Technological forecastsareconcernedwithratesof technological progress, whichcanresult inthebirthof excitingnewproductsrequiring newplantsandequipment 3.Demand forecasts are projections of a companys products or servicesit alsocalledsalesforecasts 11/27/2013 5 2.Types of Forecasting in Decision Making: Marketing Demand forecasting of products Forecast of market share Forecasting trend in prices Production Forecast of Materials requirements Trend in material and labor costs Maintenance requirements Plant capacity 11/ 27/ 2013 6 4 Finance Forecast of Cash flows Rates of expenses Revenues Personnel Forecast of Number of workers in each category Labor turn over absenteeism 11/ 27/ 2013 7 Seven steps in the forecasting system 1) Determine the use of the forecast 2) Select items to be forecast 3) Determine the time horizon of the forecast 4) Select the forecasting models 5) Gather the data needed to make the forecast 6) Make the forecast 7) Validate and implement the results 11/27/2013 8 5 Demand Patterns Forecasting is based on the pattern of events in the past. A pattern maysolelyexist asafunctionof time. Suchapatterncanbeidentified directly fromhistorical data. Another pattern consists of relationship between two or more variables. So it is important to understand the most commondemandpatterns. Historical pattern(Stationarypattern): Thisexistswhenthere is no trend in data and when the mean value does not change over time. Example: Products with stable sales. Moving, weighted moving and exponential smoothingetc approachesareused. 11/ 27/ 2013 9 Fig: Historical pattern (Stationary pattern) Fig: Level demand pattern time time F o r e c a s t
v a r i a b l e F o r e c a s t
v a r i a b l e Seasonal demand pattern: This demand pattern exists when the series fluctuates according to some seasonal factor. The season may be months, quarters, weeks, etc. sale of refrigerators, sale of refrigerators, sale of soft drink, sale of wool items, etc. Cyclical pattern: In this type of pattern the length of a single cycle is longer than a year. This cycle does not repeat at constant intervals of time. The best examples are the prices of some metals, gross national product, etc. 11/ 27/ 2013 10 F o r e c a s t
v a r i a b l e time Fig: seasonal demand pattern F o r e c a s t
v a r i a b l e time Fig: seasonal demand pattern 6 Trend pattern: This type of pattern exists when there is an increase or decrease in the value of the variable over time. The examples are sales many products, stock prices, business and economic indicators. 11/ 27/ 2013 11 F o r e c a s t
v a r i a b l e time Fig: Cyclical pattern F o r e c a s t
v a r i a b l e time Fig: Trend pattern Forecasting Models / Approaches 1. Qualitativeforecasts: It use subjective approaches. These are useful where no data is available and are useful new products. It forecasts that incorporate such factors as the decision makers intuition, emotions, personal experiencesandvaluesystems. 2. Quantitativeforecasts: It is based on historical data. These are more accurate and computerscanbeusedto speeduptheprocess. It forecasts that employone or moremathematical models that rely onhistorical dataandcausal variablestoforecast demand. 11/27/2013 12 7 Qualitative methods 1. Jury of executive opinion: A forecastingtechniquethat takes the opinion of a small group of high-level managers and results in a groupestimateof demand. 2. Delphi methods: A forecasting technique using a group processthat allowsexpertstomakeforecasts. 3. Sales forces composite: A forecasting technique based on salespersonsestimatesof expectedsales. 4. Consumer market survey: A forecastingmethod that solicits input from customers or potential customers regarding future purchasingplans. 11/27/2013 13 Quantitative Methods Timeseriesmodels: Aforecastingtechnique that usingaseries of past datepointstomakeaforecast. Associative (or causal) models: such as linear regression, incorporate thevariablesor factorsthat might influencethequantitybeingforecast. Timeseriesmodelsaregivenbelow: 1.Naiveapproach: Aforecastingtechniquethat assumesdemandinthe next periodisequal todemandinthemost recent period. 2.Movingaverage: Aforecastingmethod that uses an average of the n most recent periodsof datatoforecast thenext period. 11/27/2013 14 8 11/27/2013 15 Month Actual shed sales 3 month movingaverage January 10 February 12 March 13 April 16 (10+12+13)/3=11.667 May 19 (12+13+16)/3=13 June 23 (13+16+19)/3=16 July 26 (16+19+23)/3=19 August 30 (19+23+26)/3=22 September 28 (23+26+30)/3=26 October 18 (26+30+28)/3=28 November 16 (30+28+18)/3=25 December 14 (28+18+16)/3=20 Month Actual shed sales 3 month Weighted moving average January 10 February 12 March 13 April 16 (.2*10+.3*12+.5*13)/1=12.1 May 19 (.2*12+.3*13+.5*16)/1= June 23 (.2*13+.3*16+.5*19)/= July 26 (.2*16+.3*19+.5*23)/1= August 30 (.2*19+.3*23+.5*26)/1= September 28 (.2*23+.3*26+.5*30)/1= October 18 (.2*26+.3*30+.5*28)/1= November 16 (.2*30+.3*28+.5*18)/1= December 14 (.2*28+.3*18+.5*16)/1= 11/ 27/ 2013 16 9 To be continue 11/27/2013 17 4.Exponential smoothing A weighted moving average forecasting technique in which data points are weighted by an exponential function. Formula : New forecast =last periods forecast + (Last periods actual demand Last periods forecast) Where is a weight, or constant , chosen by the forecaster, that has a value between 0 and 1. Now F t =F t-1 +( A t-1 F t-1 ) Where F t =New forecast F t-1 =Previous forecast, A t-1 =Previous periods actual demand and =smoothing ( or weighting) constant ( 0 1) If is not given than =2 / (n+1) [n =no of period ] 11/ 27/ 2013 18 10 Example In January , a car dealer predicted February demand for 142 Ford Mustangs. Actual February demand was 153 autos. Using a smoothing constant chosen by management of =.20, we can forecast March demand using the exponential smoothing model. Answer: New forecast (for March demand) =142 +.2( 153-142) =142+2.2 =144.2 Thus, the March demand forecast for Ford Mustangs is rounded to 144. Note: the smoothing constant, , is generally in the range from .05 to .50 for business applications. 11/ 27/ 2013 19 Weight Assigned to 11/ 27/ 2013 20 Smoothing constant Most recent period ) 2 nd Most recent period (1-) 3 rd Most recent period (1-) 2 4 th Most recent period (1-) 3 5 th Most recent period (1-) 4 =0.1 0.1 0.09 0.081 0.073 0.066 =0.5 0.5 0.25 0.125 0.063 0.031 11 5.Linear Regression Regression means dependenceand involves estimating the value of a dependent variable Y, from an independent variable X. In simple regression, only one independent variable is used, whereas in multiple regression two or more independent variable are involved. The simple regression takes the following form Y =a +bX Where Y - dependent variable X independent variable a intercept b slope (trend) 11/ 27/ 2013 21 These are represented in the following graph: [The model for multiple linear regression is shown below: Y =a +b 1= X 1 +b 2 X 2 +b 3 X 3 +b i X i ++b n X n ] The formulas to compute the constants of the model are given below b =Which is written by hand. a =Which is written by hand. And n is the umber of pairs of observations made. 11/ 27/ 2013 22 Y X a b Fig: Graph showing simple regression 12 Exampl e A firm believes that its annual profit depends on its expenditures for research. The information for the preceding six years is given below. Estimate the profit when the expenditure is 6 units. 11/ 27/ 2013 23 Year Expenditure for research (X) Annual profit (Y) 1 2 20 2 3 25 3 5 34 4 4 30 5 11 40 6 5 31 7 6 ? Solution: X bar =30/6 =5 and Y bar =180/6 =30 11/ 27/ 2013 24 Year X Y XY X 2 1 2 20 40 4 2 3 25 75 9 3 5 34 170 25 4 4 30 120 16 5 11 40 440 121 6 5 31 155 25 Total 30 180 1000 200 13 11/ 27/ 2013 25 b =2 And a =20 The model is : Y =a +bX =20 +2X The profit when expenditure is 6 units: Y =20 +2 * 6 =32 Measure of forecasting accuracy Some common measures are inevitable to measure the accuracy of a forecasting technique. This measure may be an aggregate error (deviation) of the forecast values from the actual demands. The different types of errors which are generally computed are as presented below. 1. Mean Absolute Deviation (MAD) 2. Mean Square Error (MSE) 3. Mean Forecast Error (MFE) 4. Mean Absolute Percent Error(MAPE) The formula for error is given below. e t =D t F t Where D t =demand for the period t. F t =forecast demand for the period t, and e t =forecast error for the period t. 11/ 27/ 2013 26 14 Mean Absolute Deviation (MAD) 11/ 27/ 2013 27 Mean Square Error (MSE) 11/ 27/ 2013 28 15 Mean forecast error (MFE) 11/ 27/ 2013 29 Example Suppose that a forecast of 165 units had been made for the demand in every period for the data given in table. The calculation of these errors are as shown in the last two columns of table. 11/ 27/ 2013 30 t Demand D t Forecast F t Deviation D t - F t Absolute Deviation D t - F t Squared Error (D t F t ) 2 1 150 165 -15 15 225 -10 10 2 160 165 -5 5 25 -3.125 3.125 3 165 165 0 0 0 0.00 0.00 4 175 165 +10 10 100 5.710 5.71 5 180 165 +15 15 225 8.330 8.330 Total +5 45 575 27.165 16 MAD =45/5 =9 MSE =575/5 =115 MAPE =27.165/5 =5.433% MFE =+5 / 5 =+1 11/ 27/ 2013 31 Thanks to all 11/ 27/ 2013 32