Evans Analytics2e Ppt 09

Chapter 9 Forecasting Techniques

Forecasting Forecast ing Techniques Techniques  



Managers require good forecasts of future events. Business Analysts Analysts may choose from a wide range of forecasting techniques to support decision making. Three major categories of forecasting approaches: 1. ualitative and judgmental techniques !. "tatistical time#series models $. %&planatory'causal models

Qualitative and Judgmental Forecasting 







ualitative ualitative and (udgmental techniques rely on e&perience and intuition. intuition. They are necessary when historical data is not availa)le or when predictions are needed far into the future. The historical analogy approach o)tains a forecast through comparative analysis with prior situations. The Delphi method questions an anonymous panel of e&perts !#$ times in order to reach a convergence of opinion on the forecasted varia)le.

Example 9.1 !redicting the !rice o" #il  

  



%arly 1*++ , oil price was a)out -!! a )arrel Mid#1*++ , oil price pr ice dropped to -11 -11 a )arrel )ecause of oversupply oversupply high production in non# /0% regions and lower than normal demand 2n the past /0% would raise the price of oil. 3istorical analogy would forecast a higher price. 3owever the price continued to drop even though /0% agreed to cut production. 3istorical analogies cannot always account for current realities4

$ndicators and $ndexes 



$ndicators are measures that are )elieved to

influence the )ehavior of a varia)le we wish to forecast. 2ndicators are often com)ined quantitatively into an index a single measure that weights multiple indicators thus providing a measure of overall e&pectation. ◦

%&le: 5ow (ones 2ndustrial Average

Example 9.% Economic $ndicators 

650 76ross 5omestic 0roduct8 measures the value of all goods and services produced. 





650 rises and falls in a cyclic fashion.

9orecasting 650 is often done using leading indicators 7series that change )efore the 650 changes8 and lagging indicators 7series that follow changes in the 6508 indicators. %&les eading - formation of )usiness enterprises # percent change in money supply 7M18 agging # )usiness investment e&penditures # prime rate # inventories on hand

Example 9.& 'eading Economic $ndicators 





 An Index of Leading Indicators was developed )y the 5epartment of ommerce. This inde& is related to the economic performance is availa)le from www.conference#)oard.org. 2t includes measures such as: # average weekly manufacturing hours # new orders for consumer goods # )uilding permits for private housing # ";0 num)er of periods t  > 1 ! ? T 

Time series generally have components such as: # random )ehavior  # trends 7upward or downward8 # seasonal effects # cyclical effects (tationary time series have only random )ehavior.  A trend is a gradual upward or downward movement of a time series.

Example 9.* $denti"ying Trends in a Time (eries 

The Energy Production & Consumption ◦

6eneral upward trend with some short downward trends@ the time series is composed of several different short trends.

(easonal E""ects 

 A seasonal e""ect is one that repeats at fi&ed intervals of time typically a year month week or day.

Cyclical E""ects 

Cyclical e""ects descri)e ups and downs over a

much longer time frame such as several years.

Forecasting )odels "or (tationary Time (eries  

Moving average model %&ponential smoothing model ◦

These are useful over short time periods when trend seasonal or cyclical effects are not significant

)oving +verage )odels 



The simple moving average method is a smoothing method )ased on the idea of averaging random fluctuations in the time series to identify the underlying direction in which the time series is changing. The simple moving average forecast for the ne&t period is computed as the average of the most recent k  o)servations. ◦

arger values of k  result in smoother forecast models since e&treme values have less impact.

Example 9., )oving +verage Forecasting 



The Tablet Computer Sales data contains the num)er of units sold over the past 1 weeks.

Three#period moving average forecast for week 1+:

(preadsheet $mplementation o" )oving +verage Forecasting

Excel Moving Average Tool 

Data nalysis options

e do not recommend using the chart or error options )ecause the forecasts generated )y this tool are not properly aligned with the data

Example 9.- )oving +verage Forecasting ith XLMiner  



"elect Smoot!ing  from the Time Series group and select "o#ing #erage %nter the data range and move the time varia)le and dependent varia)le to the )o&es on the right. %nter the interval 7k 8.

Examnle 9.- Continued 

 $L"iner 

results

Error )etrics and Forecast +ccuarcy 









Mean a)solute deviation 7MA58

Mean square error 7M"%8

Coot mean square error 7CM"%8

Mean a)solute percentage error 7MA0%8

Example 9./ 0sing Error )etrics to Compare )oving +verage Forecasts 





Tablet Computer Sales data

!# $# and D#period moving average models !#period model ahs lowest error metric values

Exponential (moothing )odels 



(imple exponential smoothing model:

where % t' is the forecast for time period t  E 1 % t  is the forecast for period t  t  is the o)served value in period t  and α  is a constant )etween = and 1 called the smoothing constant. To )egin set % ' and % (  equal to the actual o)servation in period 1 '.

Example 9.9 0sing Exponential (moothing to Forecast Tablet Computer Sales

9orecast for week $ when

α

> =.: 71 , =.87++8 E 7=.87DD8 >