Forecasting Principles and methods

Advanced models of time series analysis and these are known as exponential smoothing models. The name explanation smoothing is derived from weighted moving average method. We used to assign weights to the demand data of previous periods now if the weights are decreasing exponentially it becomes exponential smoothing model. It is basically type of weighted moving average method so in this exponential smoothing method we assign weights but these weights are decreasing in the exponential order from the present period to the past periods. In this model we use word smoothing is the characteristics of our past data we use some data and in that we saw that there is no smooth curve available for historical data even in the case of horizontal demand data the curve is like given below:

there are fluctuations and jig-Jack movement in my past data. It is argued that these jig jag movements are around a baseline (as in fig) and around baseline this jig Jack movement is taking place. When we are using this word smoothing, so we are trying to smooth this jig jag line and touch the baseline. It can touch the baseline. Whenever actual demand data will be there, there will be some kind of fluctuations and these fluctuations are difficult to control of anybody you cannot control these fluctuations but what we can do we match the baseline with smoothing curve. The weight are reducing in the exponential order and the second is to smooth the fluctuations of this actual demand line so that actual demand line can coincide with baseline. These types of models are very efficient and the power of these models that computation becomes much easier.

When to use Exponential smoothing:

1.      When you are forecasting for a large number of items.

2.      The forecasting horizon is relatively short.

3.      There is little outside information available about the cause and effect.

4.      Some effort in forecasting is desired. Effort is measured by both a method’s ease of application and by the computational requirement.

5.      Updating of the forecast as new data becomes available is easy.

6.      It is desired that the forecast is adjusted for randomness and tracks, trends and seasonality.

How we apply the method for smoothing and remove randomness the formula for updating the base value is:

New Base = Previous Bases + α(New Demand – Previous Base)

 Smoothing Constant, α must be between 0 to 1

A large α provides a high impulse response forecast.

A Small α provides a low impulse response forecast.

the values of α can range from 0 to 1. Your experience and there are some software help also available which can help us what is the right value of alpha.

There are two cases of α becomes 0 and then New base equal to previous base.

In that situation totally discard that fluctuations and in that situation I am taking α=0