What is Advanced Analytics?
Almost every day I hear about “when machines will replace man,” that the day-to-day repetitive tasks will end, that robots will do everything for us. Have you heard about it? Yeah, I guess so!
And what’s the problem with all this? None, from my point of view. I think that in fact, all that is coming is going ‘more for our good than against us’. Have you ever thought that you will no longer have to cook for duty, wash your clothes or even drive? Is not fantastic? We will be able to choose what to do and when to do it.
Many still wonder: what about us? What will happen about our jobs? Will life have no more joy? I would answer: yes, we will enjoy our lives and a lot. But in another way, different ways of enjoying life will appear, you can write that down on a piece of paper;)
Hey Sig, What does this have to do with the subject? It’s all about! It turns out that when we are talking about Advanced Analytics we are thinking about future, predictions, and hypotheses about what has not happened yet.
OK, imagine a timeline about your company’s history to this day. This view can be given by Business Intelligence (BI) applications, but from this point on, from today to the future, only Advanced Analytics (AA) can help you. This means that BI sees past and present while AA looks to the future.
Look at this image released by EMC. There, you can find the relationship between time and value for the business. AA is right in the area where Data Science comes in. That is, the Data Scientist is the professional that works with AA seeking to add more value to the business looking not only for the past and present but for the future.
While BI seeks to respond “What happened?” by displaying Reports and Charts, AA is concerned with answering questions such as “What is likely to happen?” or “What should I do?” in the areas of Predictive and Prescriptive analytics.
The image below, by Garnter, complements this explanation, however in this chart instead of plotting time versus value for the business, Gartner plotted difficulty versus value. This comparison is very interesting since, in fact, making predictions requires a more detailed knowledge of Statistics and Mathematics.
The creation of future scenarios and forecasts requires a great understanding of the Business and about Statistics/Mathematics. It because giving to the Managers an incorrect forecast can turn into a real headache. So, be careful and study that before start!
I have seen many professionals perform a linear regression on a negligible mass of data and intended to generate forecasts. Or even, generating regressions incompatible with the behavior of the cloud of points.
My suggestion is to start with Time Series. Understanding “lines” behavior, seasonality, stationary... Seek to study different algorithms that can be used to perform regression, not only linear. Go beyond!
Much more important than knowing how to use a library or even presenting a beautiful chart with forecast lines, you must know what does the data that these libraries return actually “mean”.
I suggest starting framing the problem. Understanding the questions that Managers want to answer, What their indicators are (KPIs) and Where they want to go.
With a good start, the possibility of future failure is reduced ;)
In the next articles, we’ll talk more about this fantastic universe of forecasting.
Good Forecasts!
