Forecasting using Predictive Functions in Lumira 1.31
Forecasting is the process of making predictions of the future based on past and present data and analysis of trends – Wikipedia
As an extension to the data discovery aspects of Lumira, we can perform basic predictions or forecasts using the Predictive Calculation option.
There are three important forecast methods available in Lumira,
1) Linear Regression
Simple Linear Regression fits a straight line through the set of n points in such a way that makes the sum of squared residuals of the model (that is, vertical distances between the points of the data set and the fitted line) as small as possible. – Wikipedia
2) Triple Exponential Smoothing
Exponential smoothing is a rule of thumb technique for smoothing time series data, particularly for recursively applying as many as three low-pass with exponential window functions. It is an easily learned and easily applied procedure for approximately calculating or recalling some value, or for making some determination based on prior assumptions by the user, such as seasonality. – Wikipedia
3) Using SAP Predictive Analytics
This would be one of the most accurate methods of forecasting. But the accuracy depends on the size of the historical data, the larger the size of the data the better. If the number of data points is lesser than 16 or if the pattern is not visible we can use option 2 instead.
For a more advanced forecasting and leveraging other predictive algorithms, I would use a tool like SAP Predictive Analytics.
Certain conditions do need to be satisfied before forecasting,
This is the error message suggesting the changes that you would need to leverage the predictive capabilities.
To start with basic prediction, we can use a Line Graph across Month as the time dimension for more data points. We can extend this concept across other relevant charts or measures.
Under the measures we can see the option for Predictive Calculation,
We can select the Forecast method and the interval of prediction as follows,
The forecasted values will be visible on the chart and can be accessed under the measure axis. The following chart shows predictions based on the Forecast methods,
The method of forecasting depends on the situation. I would use smoothing method for month-month forecast of sales numbers and use regression for a more granular level analysis involving variables which are positively or negatively tied together.