< Previous | Next > |

Time Series data holds variations, to reduce or cancel the effect of these variations we use smoothing techniques.

Cases where time series data is stable means has no trend, seasonal or cyclical effects, we can use smoothing methods to average out the irregular component of time series.

Common Smoothing Methods are:

- Moving Average
- Weighted Moving Average
- Centered Moving Average
- Exponential Smoothing

**Moving Average:**

Moving Average techniques are useful if one can assume item to be forecast will stay steady over time.

The moving averages method consists of computing an average of the most recent N observations for the series and forecasting value of the time series for the next period.

Moving Average = S ( recent N observations) / N

So, let’s start with example now

A manager of Multinational Company wants to know the demand of products for next period. He has recorded the demand of last twelve months.

The Average of data is 22.5. So manager decides to use this as the estimate of demands. But how much good is this estimate, this we can check using “Mean Squared Error” calculations for different different estimates.

In MSE (Means Squared Error), Error is the actual demand minus the estimate.

Squared is square of errors.

As per statistics, estimate with small MSE is good. So in the below screenshot estimate 22.5 have small MSE which is 25.3

< Previous | Next > |