Srraaa

Only available on StudyMode
  • Download(s) : 41
  • Published : December 13, 2012
Open Document
Text Preview
Forecasting Models
NMIMS
Forecasting techniques

Qualitative modelstime series modelscausal models
1.Delphi method1.moving averages1.regression analysis
2.Opinion poll2.exponential smoothing2.multiple regression
3.Historical Analogy 3.econometric models 4.Field Surveys
5.Business barometers
6.Extrapolation Technique
7.Input-Out put Analysis
8.Lead Lag Analysis
9.Sales force composites
10.Consumer Market survey
Simple Average Method
The historical data is used for extrapolating and forecasting. Either simple averages or moving averages could be used. In simple average method, for establishing the trend, the data is divided in two parts, and from the change, the trend is established. From seasonal averages, “seasonalizing indices” for seasons are calculated. For forecasting purposes, the forecast based on trend analysis is first calculated, and this forecast is then “seasonalized” by using seasonalizing factor. While using seasonal data, first data is “deseasonalized” by dividing it by the seasonalizing factor; then the forecast is made by trend analysis; and finally the forecast is “seasonalized” by multiplying by the seasonalizing factor. In fact this procedure of “deseasonalize - forecast - seasonalize” can be used with any method of forecasting. Moving Average Method

In moving average method, a moving average of suitable periods is used for developing the forecast for the next period. Time series methods generally require large amount of historical data to be available at hand. They are suitable for products, the demand for which is sustained, and is not prone to change due to fashions and change in public tastes. Exponential smoothing methods

These methods are suitable for situations where the immediate past data is more relevant than the old data for making the forecast. The demand for consumer goods items, which is subject to change due to changes in public tastes and for the products having a short life cycle, is forecasted by this method. It is commonly used by supermarkets for deciding the quantity to be kept on shelf every buying period. Single Stage Exponential Smoothing

The forecast for the coming period(Ft+1) is based on the demand for the current period (Dt ) and the forecast for the current period (Ft), using the formula:
(Ft+1) = (Dt ) + (1 - )(Ft),
where is called the smoothing constant and can have any value between 0 and 1. For deciding value of , the following heuristic formula can be used: = 2/(n + 1)
where n is the number of periods used for smoothing. Where data for more number of periods is available, it can be used to find out by trial and error, the value of which minimizes mean absolute deviation in calculating the forecast. Two-Stage Exponential smoothing

Where a changing trend is also expected, two-stage exponential smoothing is used. In this case, the exponential smoothing procedure is also used for smoothing error between the forecasted trend and the actual trend. Regression and correlation Methods

Method of Least Squares
The statistical method of least squares can also be used to find the line of best fit in the available historical data, which on extrapolation would give trend as well as the forecast for the next period. Regression analysis

When the demand can be linked with some other random variable, the demand can be forecasted by establishing the relationship between demand and the random variable, by regression analysis. The demand can then be forecasted from the known value of the variable during the period for which the forecast has to be made.

Problems :
1) The accompanying table shows demand for the past 16 months: -------------------------------------------------------------------------------------------------------- Month 1 2 3 4 5 6 7 8 Demand 8.9 9.5...
tracking img