Forecasting Dr. Richard Jerz 1 1
Learning Objectives Describe why forecasts are used and list the elements of a good forecast. Outline the steps in the forecasting process. Describe at least three qualitative and three quantitative forecasting techniques, and their advantages and disadvantages. Describe two measures of forecast accuracy. Identify the major factors to consider when choosing a forecasting technique. 2 2
Making Decisions If we know exactly what will happen in the future, operational decisions would be easy. Problem we don t know the future. 3 3
FORECAST A statement about the future Used to help managers Plan the system Operate the system?? Forecasting time horizons Short-range Medium-range Long-range 5 4
Forecasts Underlying basis of all business decisions! 6 5
Department Uses of Forecasts Accounting Cost/profit estimates Finance Cash flow and funding Human Resources Hiring/recruiting/training Marketing Pricing, promotion, strategy MIS IT/IS systems, services Operations Schedules, MRP, workloads Product/service design New products and services 7 6
Forecasts Characteristics Assumes causal system and system stability past ==> future Forecasts rarely perfect because of randomness Forecasts more accurate for groups vs. individuals Forecast accuracy decreases as time horizon increases Statistics & math are often used 11 7
Elements of a Good Forecast Timely Reliable Accurate Meaningful Written Easy to use 12 8
Seven Steps in Forecasting 1. Determine the use of the forecast 2. Select the items to be forecasted 3. Determine the time horizon of the forecast 4. Select the forecasting model(s) 5. Gather the data 6. Make the forecast 7. Validate and implement results 13 9
Types of Forecasts Judgmental - uses subjective inputs Time series - uses historical data assuming the future will be like the past Associative models - uses explanatory variables to predict the future 14 10
Judgmental Forecasts Qualitative Executive opinions Sales force composite Consumer surveys Outside opinion Opinions of managers and staff Delphi method 15 11
Time Series Forecasts Quantitative Trend - long-term movement in data Seasonality - short-term regular variations in data Cycles wavelike variations, usually more than one year s duration Irregular variations - caused by unusual circumstances Random variations - caused by chance 17 12
Forecast Pattern Examples 18 13
Most Common Quantitative Methods Naïve forecasts Moving average Weighted moving average Exponential smoothing Trend analysis Time-Series Models 20 14
Naïve Forecast The forecast for any period equals the previous period s actual value. Can be adjusted with trend Ft = At 1 ( + trend) 21 15
Naïve Forecasts Advantages Simple to use Virtually no cost Data analysis is nonexistent Easily understandable Can include trend and seasonality considerations Disadvantages Cannot provide high accuracy Can be a standard for accuracy 22 16
Moving Average A technique that averages a number of recent actual values, updated as new values become available. Moving average = demand in previous n periods n F t = A +... + At 2 At n t n + 1 23 17
Simple Moving Average Characteristics Must choose the number of periods, n Only use the most recent n periods Gets rid of old data Premise: newer data is more indicative of what the future will be The bigger the n, the less sensitive the forecast The smaller the n, the more reactive the forecast Does not forecast trend well 24 18
Moving Average: Periods 47 45 43 41 39 37 35 Actual 1 2 3 4 5 6 7 8 9 10 11 12 MA5 MA3 25 19
Weighted Moving Average More recent values in a series are given more weight in computing the forecast. Weighted moving average = (weight for period n) n x (demand in period n) n weights wn At n +... + wn 1At 2 + w1 At 1 Ft = w 27 20
Weighted Moving Average Characteristics Weights are usually a percent Sum of the weights equals 1, or 100% More complex than simple moving average More responsive to most recent events Weights based on experience and intuition Better (than simple moving average) at forecasting trend 28 21
Exponential Smoothing Premise -- The most recent observations might have the highest predictive value. Therefore, we should give more weight to the more recent time periods when forecasting. New forecast = last period s s forecast + α (last period s s actual demand last period s s forecast) F t = F + α t 1 t 1 t ( A F 1) 29 22
Exponential Smoothing Characteristics A form of weighted moving average Requires a smoothing constant (alpha, α) that ranges from 0 to 1 The larger alpha, the more reactive the forecast model. Good at forecasting trend Involves little record keeping of past data 30 23
Trend Analysis Linear trends - Fitting a trend line to historical data points to project into the medium-tolong-range Linear trends can be found using the least squares technique ^ y ^y ^ = a + bx where y = computed value of the variable to be predicted (dependent variable) a = y-axis y intercept b = slope of the regression line x = the independent variable 31 24
Least Squares Method Values of Dependent Variable Actual observation (y value) Deviation 3 Deviation 1 Deviation 2 Deviation 7 Deviation 5 Deviation 6 Deviation 4 Trend line, y ^ = a + bx 32 Time period 25
Equations to Calculate the Regression Variables y ^y ^ = a + bx Σxy b = xy - nxy Σx 2 - nx 2 33 a = y - bx 26
Least Squares Example Time Electrical Power Year Period (x) Demand x 2 xy 1999 1 74 1 74 2000 2 79 4 158 2001 3 80 9 240 2002 4 90 16 360 2003 5 105 25 525 2004 6 142 36 852 2005 7 122 49 854 Σx = 28 Σy = 692 Σx 2 = 140 Σxy = 3,063 x = 4 y = 98.86 Σxy - nxy 3,063 - (7)(4)(98.86) b = = = 10.54 Σx 2 - nx 2 140 - (7)(4 2 ) a = y - bx = 98.86-10.54(4) = 56.70 34 27
Power demand Least Squares Example Trend line, 160 y ^ = 150 56.70 + 10.54x 140 130 120 110 100 90 80 70 60 50 1999 2000 2001 2002 2003 2004 2005 2006 2007 Year 35 28
Trend Line Characteristics Great, if data has trend Variations around the line are assumed to be random If trend is not linear, cannot use trend line Trend could be curves, but math becomes more difficult Plot the data to insure a linear relationship Be careful forecasting far beyond the database 36 29
More Advanced Techniques Associative forecasting Predictor variables not time Multiple linear regression More than one predictor variable 37 30
Forecast Accuracy We generally do this by selecting the model that gives us the lowest forecast error Error - difference between actual value and predicted value Mean absolute deviation (MAD) Average absolute error Mean squared error (MSE) Average of squared error Correlation coefficient for trend line 38 31
Mean Error Difference between actual and forecast Absolute value used ( Actual Mean Error = n Forecast) 39 32
Mean Squared Error The error is squared Variance Mean Squared Error ( Actual Forecast) = n 1 2 40 33
Correlation and Linear Regression Are two sets of data related? Correlation analysis Is the relationship linear? Linear regression analysis How strong is the relationship? Correlation Coefficient of Correlations (denoted by r). Values range from -1 to +1 Values close to 0.0 indicate a weak correlation Negative values indicate an inverse relationship 41 34
Correlation 42 35
Correlation 43 36
Correlation Coefficient Equation to Calculate r 44 37
Choosing a Forecasting Method Cost Accuracy Timely Understandable Serves purpose In writing 46 38
Operations Strategy Forecasts are the basis for many decisions Work to improve short-term forecasts Accurate short-term forecasts improve Profits Lower inventory levels Reduce inventory shortages Improve customer service levels Enhance forecasting credibility Consider accountability of person making the forecast 47 39