forecasting6
DESCRIPTION
Forecasting TechniquesTRANSCRIPT
Overview of MethodsOverview of Methods
Quantitative TechniquesQuantitative Techniques
Moving AverageMoving AverageTrend AnalysisTrend AnalysisExponential SmoothingExponential SmoothingARIMA modelsARIMA modelsEconometric modelsEconometric models
Moving AverageMoving Average
A simple average of the previous X A simple average of the previous X months/yearsmonths/years
A six-month moving average forecast is an A six-month moving average forecast is an average of the previous six monthsaverage of the previous six months
““I always avoid prophesying I always avoid prophesying beforehand because it much beforehand because it much better to prophesy after the better to prophesy after the
event has already taken place.”event has already taken place.”
Winston ChurchillWinston Churchill
Moving Average – When To UseMoving Average – When To Use
Extremely “noisy” or little dataExtremely “noisy” or little data
Time constraintTime constraint
Degree of accuracy not importantDegree of accuracy not important
Moving Average - AdvantagesMoving Average - Advantages
Extremely simpleExtremely simple
Easy to implementEasy to implement
Moving Average - DisadvantagesMoving Average - Disadvantages
Not accurate; slow adjustment to changes Not accurate; slow adjustment to changes in datain data
Misses turning pointsMisses turning points
All history is created equalAll history is created equal
Moving Average ExampleMoving Average Example
Trend RegressionTrend Regression
A straight ( or curved) line drawn through A straight ( or curved) line drawn through historical datahistorical data
““taking a ruler through your data”taking a ruler through your data”
““The best qualification of a The best qualification of a prophet is to have a good prophet is to have a good
memory.”memory.”
Marquis of HalifaxMarquis of Halifax
Trend Regression – When To UseTrend Regression – When To Use
Steady rise or decline in dataSteady rise or decline in data
Time or software constraintTime or software constraint
Need easy explanationNeed easy explanation
Little dataLittle data
Trend Regression - AdvantagesTrend Regression - Advantages
Very simpleVery simple
Can be done in ExcelCan be done in Excel
Trend Regression – DisadvantagesTrend Regression – Disadvantages
Assumes future is exactly like past (prices, Assumes future is exactly like past (prices, economy, etc.)economy, etc.)
All history is created equalAll history is created equal
One bad data point can greatly affect One bad data point can greatly affect forecastforecast
Trend Regression ExampleTrend Regression Example
Exponential SmoothingExponential Smoothing
SimpleSimple
Double (Brown) or HoltDouble (Brown) or Holt
WintersWinters
““A good forecaster is not A good forecaster is not smarter than everyone else, he smarter than everyone else, he merely has his ignorance better merely has his ignorance better
organized.”organized.”
C. W. J. GrangerC. W. J. Granger
Simple Exponential SmoothingSimple Exponential Smoothing
Weighted average of past values with Weighted average of past values with exponentially decreasing weightsexponentially decreasing weights
Forecast this month equals last month’s Forecast this month equals last month’s forecast plus a proportion of the forecast forecast plus a proportion of the forecast error last montherror last month
Simple Exponential Smoothing – Simple Exponential Smoothing – When To UseWhen To Use
Stationary data with no trend or Stationary data with no trend or seasonalityseasonality
Double (Brown) or Holt Exponential Double (Brown) or Holt Exponential SmoothingSmoothing
Smooth the smoothed data with a Smooth the smoothed data with a weighted average of past values with weighted average of past values with exponentially decreasing weightsexponentially decreasing weights
Changes linearly with time (like linear Changes linearly with time (like linear regression) with recent data given more regression) with recent data given more weightweight
Double (Brown) or Holt Exponential Double (Brown) or Holt Exponential Smoothing – When to UseSmoothing – When to Use
Data with a trend but no seasonalityData with a trend but no seasonality
Winter’s Exponential SmoothingWinter’s Exponential Smoothing
Deseasonalize data, then find trend, then Deseasonalize data, then find trend, then smoothsmooth
Winter’s Exponential Smoothing – Winter’s Exponential Smoothing – When to UseWhen to Use
Data with trend and seasonalityData with trend and seasonality
Exponential Smoothing AdvantagesExponential Smoothing Advantages
Somewhat simpleSomewhat simple
Recent data given more weightRecent data given more weight
Fairly good accuracy for short-term Fairly good accuracy for short-term forecastsforecasts
Software can automate processSoftware can automate process
Exponential Smoothing - Exponential Smoothing - DisadvantagesDisadvantages
Requires forecasting softwareRequires forecasting software
Bad data in recent month can cause great Bad data in recent month can cause great error in forecasterror in forecast
Less accurate for medium to long-term Less accurate for medium to long-term forecastsforecasts
Assumes history is like (recent) historyAssumes history is like (recent) history
Exponential Smoothing ExampleExponential Smoothing Example
ARIMA (Box-Jenkins) ModelsARIMA (Box-Jenkins) Models
AAutoutoRRegressive egressive IIntegrative ntegrative MMoving oving AAverageverage
AutoregressiveAutoregressive – future values depend on – future values depend on previous values of the dataprevious values of the data
Moving averageMoving average – future values depend on – future values depend on previous values of the errorsprevious values of the errors
IntegratedIntegrated – refers to differencing the data – refers to differencing the data
““An unsophisticated forecaster An unsophisticated forecaster uses statistics as a drunken uses statistics as a drunken man uses lamp-posts – for man uses lamp-posts – for
support rather than illumination”support rather than illumination”
- after Andrew Lang- after Andrew Lang
ARIMA (Box-Jenkins) Models – ARIMA (Box-Jenkins) Models – When to UseWhen to Use
Stable data that has regular correlationsStable data that has regular correlations
ARIMA ( Box-Jenkins) Models - ARIMA ( Box-Jenkins) Models - AdvantagesAdvantages
Outperforms exponential smoothing on Outperforms exponential smoothing on homogenous and stable datahomogenous and stable data
Software can automateSoftware can automate
Sounds impressiveSounds impressive
ARIMA (Box-Jenkins) Models - ARIMA (Box-Jenkins) Models - DisadvantagesDisadvantages
Requires softwareRequires software
Needs a minimum of 40 data pointsNeeds a minimum of 40 data points
Complicated to understandComplicated to understand
ARIMA (Box-Jenkins) Models ARIMA (Box-Jenkins) Models ExampleExample
Econometric ModelsEconometric Models
Relates data series to explanatory Relates data series to explanatory variablesvariables
Economists build demand models which Economists build demand models which relaterelate– Price, competition, income, population, etc.Price, competition, income, population, etc.
““An economist is an expert who An economist is an expert who will know tomorrow why the will know tomorrow why the
things he predicted yesterday things he predicted yesterday didn’t happen today.”didn’t happen today.”
Evan EsarEvan Esar
Econometric Models – When to Econometric Models – When to UseUse
Important to understand marketImportant to understand market
Influences on product demand are Influences on product demand are changingchanging
Historically more acceptable in regulationHistorically more acceptable in regulation
Econometric Models - AdvantagesEconometric Models - Advantages
Can give price elasticityCan give price elasticity
Formally integrates economic impactFormally integrates economic impact
Permits varied assumptions, i.e., “what if?”Permits varied assumptions, i.e., “what if?”
Forces you to make assumptions explicitForces you to make assumptions explicit
Methods to deal with short timeMethods to deal with short time
Econometric Models - Econometric Models - DisadvantagesDisadvantages
Large data gatheringLarge data gathering
Expertise to buildExpertise to build
Requires forecasts of explanatory Requires forecasts of explanatory variablesvariables
Not always best forecasting techniqueNot always best forecasting technique
Econometric Models ExampleEconometric Models Example
Res. DA ModelRes. DA Model
#DA calls per person = 4.49-0.18*Price #DA calls per person = 4.49-0.18*Price +1.04*Income per +1.04*Income per person+0.00016*Timetrendperson+0.00016*Timetrend
SummarySummary
Graph dataGraph data
Choose appropriate technique forChoose appropriate technique for– OutputOutput– TimeTime– DataData
Know advantages and disadvantagesKnow advantages and disadvantages