Resources

Forecasting and Time Series Analysis

Every business requires predicting the future on the basis of historical data. The prediction needs to be close to the actual behavior so that you can be benefited out of your prediction. You need to optimize the inventory using past sales record and several other business examples are there.

The increasing complexity of the business environment these days requires you to be a step ahead, implies that every organization needs to know the future values of their key decision variables. Forecasting is essential to make reliable and accurate estimates of the future.

Forecasting Methods


Here we will mainly deal with Quantitative method of Forecasting.

Time Series Analysis: Here we assume that future is going to look like the past. We need to decompose the Time series pattern into Trend, Cycles, Seasonal and Irregular behavior.

Trend: Steady tendency of either upward or downward movement in the average value of forecast variable y over time.

Cycles: An upward and downward movement in the variable about the trend line over a time period is called cycles.

Seasonal: Fluctuation repeated in a year.  Special case of cyclical movement is called seasonal.

Irregular:  Not predictable.

We need to decompose the various components of Time Series components which produce variations in the Time Series and after isolating them, analyzing the factors independently.

Two common models which are used to analyze the effect of Time Series components are Multiplicative and Additive model.

Multiplicative Model       Y = T * C * S * I

Additive Mode                  Y = T + C + S + I

TIP:  In additive Model the components can have both negative and positive values. Also we assume components to be independent of each other. However in real  - life time series analysis this does not hold true.
Now you must have got some idea of how the time series analysis is approached. We will move now to various methods of time series methods.

Freehand Method: After plotting the trend of value across time just connect the points in a freehand manner. Now draw a trend line, extend the trend line to predict future. It is not much used because it is quite a subjective method.

Smoothing Methods:  The objective of this method is to smoothen out the random variables caused by irregular components. The term “moving” is used because it is obtained by summing and averaging the values from a given number of periods, each time deleting the oldest value and adding the latest value.

Plot the new points came out after taking moving average. This will help in removing the irregular components.

Exponential Smoothing Methods:

Forecasting technique which weights past data from previous time periods with exponentially decreasing importance in the forecast so that most recent data carries more weight in the moving average.

F t = α X t -1 + (1-α) Ft-1

X t -1 Actual value for the present period

Ft-1   Forecasted value of the past period

α       Smoothing constant

Approximate value of    α = 2 / (n + 1)

Trend Projection Methods:

Linear Trend Model: The method of least squares from regression analysis is used to find the trend line of best fit to a time series data.

Other trend projection models are Quadratic Trend Model and Exponential Trend Model.

Trend Line    Y = a + b x (You can use the least square method used in regression method)

Seasonal Effects:

Can compute using Ratio to Average Method   T.S.C.I / T = S.C.I

There are several other methods of calculating Seasonal Index such as Methods of Simple Average, Ratio to Moving Average Method and Link Relatives Method.

Diagrammatic representation of cyclical and trend component of time series

Watch out for more very soon!

Web link : Market Research Data Analytics Solutions provider

« back