Making decisions knowing in advance what will happen in the future is everyone's wish. And we currently have only two ways to try to predict something: by consulting crystal balls and other supernatural powers, or by using statistics and artificial intelligence that enable us to make better data-driven decisions.
The good news is that the second option is already a reality and is available on Cognos Analytics with Watson. Best of all, you don't need to understand absolutely anything about statistics or data science, since IBM has created a very easy and user friendly functionality that allows a line business user to use the forecast with just one click
What is Time Series Forecasting?
Time series forecasting is a technique that uses a machine learning or statistical approach to analyze data from the past, identifies a pattern and tries to predict a trend that allows suggesting future results
Where can the Forecast be applied in real life?
Imagine if you could predict how much you will sell this month. Or maybe if you could know how much of a specific product you need to have in stock to avoid losing sales and have a better inventory?
Okay, sales-related examples don't convince you and you need a stronger reason. So imagine if you could estimate the future value of corporate stocks at a specified time interval. Or if you can forecast the number of people infected by a virus during a pandemic like Covid? These and many other examples can be potential scenarios for using time series forecasting.
Univariate x Multivariate Forecast
Until it’s 11.2.2 release, Cognos Analytics already offered what we call univariate forecast, which means that the user can use the forecast feature considering just one factor (in addition to time) as an input parameter for Cognos Artificial intelligence to predict future results.
In the 11.2.3 release, Cognos Analytics with Watson introduced the unprecedented capability called Multivariate Time-series Forecasting. It means, that now the user can forecast using multiple parameters.
A simple and didactic use case to understand it in practice:
Imagine that a logistics company wants to predict the volume of goods that will arrive at its warehouse in the coming weeks. This company can use the simple time series forecast to analyze the volume of goods for a certain period in the past and find a trend that allows predicting what will happen in the future. Note that I'm only talking about one parameter: The volume of goods. This is the univariate forecast.
However, it would also be nice if the forecast considers other parameters, like the weather, the cost of transport, inflation, or even different seasons. In this case, as we are saying that all these other parameters can influence the result, the artificial intelligence will use the multivariate forecast
How to forecast in Cognos Analytics?
Despite the technical complexity behind how time series forecasting models works, IBM allows every Cognos Analytics user, even without any technical or statistical knowledge, to become a forecasting expert.
That's because Cognos Analytics offers an intuitive experience that automatically selects the model (from a set of nine different model types) and makes forecasting feature easy to use, even if you are not familiar with time series.
To use forecasting, you first need to have a line, bar, or column visualization on your dashboard. When data is recognized as a time series, data preparation is automated and you will see a Forecast icon in the upper right corner of this visualization.
When you click on this icon, you will see the Cognos Analytics Panel on the right side of your screen
Just enable the Forecast option and Cognos will do the rest for you. The Forecast will be displayed in the visualization as extensions of the current data.
#CognosAnalyticswithWatson