*um data scientists are creating key foundations for Machine Learning and Industry 4.0
The unbelievable machine is running at full speed: Over the past twelve months *um has been building a unique data science team around Chief Data Scientist Klaas Wilhelm Bollhoefer. The former Fraunhofer scientist Dr. Christian Thurau is using more than ten years of expertise in the field of Artificial Intelligence & Machine Learning. The rapidly expanding team currently consists of eight data scientists who collectively bring more than forty years of research and practical experience to the table. Such cumulative expertise can't be found anywhere else in Europe.
Starting with Big Data analytics
In order to gain and purposefully use knowledge based on specific, statistical processes, it's all about the "three Vs": Automatically generating and cumulating large amounts of data (Volume) from a range of sources (Variety) and quickly processing the data in up to real time (Velocity). So far, so good. However, despite its effectiveness it's still a one-way process.
Putting machine learning into practice
The next logical stage of development is to let the machine make use of the knowledge by itself. In other words, to ensure that the machine can analyze complex volumes of data based on models it creates automatically and then learn from the results. The machine generates and implements data knowledge from experience – this is artificial intelligence. An artificial system is able to learn from sample data, "recognize" patterns within it and then, following a short learning phase, apply these patterns to unknown instances. This process is used in practice within fields such as predictive analytics (the forecasting of events, losses and behaviors) and recommender systems (market basket analyses and purchase recommendations).
In 2013 analytics thought-leader Thomas H. Davenport already predicted in a Wall Street Journal article that quick modeling systems are required to keep up with the rapidly changing and growing volume of data: "Humans can typically create one or two good models a week, machine learning can create thousands."
*um is offering Machine Learning as a Service (MLaaS)
.. and is serving the huge, ever-increasing demand. When it comes to the new technologies, algorithms and programming languages that are required, many companies are unable to apply them, never mind operate them. More often than not, possessing the necessary experience and know-how proves to be the stumbling block. Therefore, reverting to standardized MLaaS offers such as Amazon AWS or Microsoft Azure ML is not the answer; companies typically don't require standard applications.
We know that each company and project has its own – usually highly specific – requirements for machine intelligence components. Therefore, our data science team is developing and modeling customized machine learning components that can be operated "as a Service" (aaS) in *um's or the customer's infrastructure and that provide their "intelligence" via defined interfaces (RESTful). Sounds simple, right? But such a service is still rarely seen, and that's why this added value is so important – not just for Industry 4.0, but much more!