Our ability to analyse data is improving exponentially and broad development perspectives are opening up. In recent times, the debate regarding data analysis has reached the general public domain. Words such as Data Science, Predictive Analytics, Advanced Analytics, Machine Learning and Artificial Intelligence are becoming universally known. These are new technologies even for many within the field of Business Intelligence but they are already being widely discussed in the media.
As a business executive, you may fall prey to stressing over the rapid technology development and wonder if you are missing out on the many new opportunities that are coming your way. My message to you is this: many of the concepts mentioned above are an evolution of each other and have been around for a long time already. Examples of early applications of predictive analytics are risk assessment of insurance portfolios and fraud and churn analysis in the telecoms industry.
However, I also want to stress this: "Something is really happening out there, now"! Opportunities to work with advanced analytics on a wider scale are now multiplying thanks to more efficient hardware, cloud solutions where computing capacity is principally unlimited and a strong increase in the rate of investment in advanced analytics by software suppliers such as Microsoft and IBM. Software tools are becoming more cost-effective and more accessible and this allows decisions, actions and outcomes to be predicted and controlled as never before.
But please note: It is not really the future we are predicting... Rafal Lukawiecki, long-time lecturer at Microsoft’s BI Days with a long background in this area – called Data Mining when Rafal started in the industry - expressed it very well a few years ago ”Customers ask me to predict the future. I can’t predict the future. But I can predict the present”. And that's the way it is. Advanced analytics is simply statistics-based models for the analysis of often very large sets of data. An example of a simple form of advanced analytics is regression analysis. But many more complex predictive models are available. However, it is important to remember, as Rafal so wisely said, that nobody can predict the future. But what we can do is work out what the most likely outcome is, based on what we know today. Note that even traditional Business Intelligence can do this to a certain degree, for example via something as simple as a Rolling Twelve Month trend analysis.
So what about the terminology – what is really in a name here? What once was called Data Mining has often been called Predictive Analytics for the last few years. Now the terms Advanced Analytics, or Data Science, are increasingly heard. In addition, the concepts of Machine Learning or Artificial Intelligence are becoming widely known. These are methods of getting software to learn a specific area well enough to carry out individual decisions and thus automate a process or help a human make the right decision. One example of this had major media impact only a week ago, when Korean Lee Sedol lost 1 - 4 against Google Deep Mind software AlphaGo, in the notoriously complex game of Go.
At Enfo, we work with more down-to-earth projects. For example, we are currently working on projects in predictive maintenance where we help industrial customers increase uptime in production by predicting when machines require maintenance and how they can be optimised. The Industrial sector is an exciting area where the Internet of Things allows connecting machinery and controlling software, giving managers access to vast amounts of additional data and the opportunity to save costs and increase output. Global competition in the industry sector means that small improvements in production efficiency can be crucial for survival.
I am convinced that we have only seen the beginning of what Data Science - Predictive Analytics - Advanced Analytics - Data Mining - Machine Learning - Artificial Intelligence - can do for humanity. Correctly used, these technologies can make us richer, healthier and happier!
But hey, don't forget the old wisdom: 80 % of the job is still having your data in good order. Without good data, there is a risk that all analyses give the wrong results, irrespective of how good a statistical model you have.
Åsa Landén Ericsson