Object Analytics Database

A 360° view is mandatory – and a relational database insufficient

Whatever your business object is (the “Patient”, “Machine”, or the “Customer”, etc.) – you are constantly collecting diverse information about it. And this much is certain: this diversity of information will even expand in the future. Therefore, it is becoming increasingly important that you can analyze your business object across all the diverse data streams related to it.

Collecting and consistently managing complex data – that’s what relational databases are great at and created for. However, to do this, you have to break an object down into its component parts and store them in different tables. Once an object is spread across many tables, it is challenging to analyze it “as a whole” – and an agile 360° view is hard to achieve.

Our Object Analytics Database – the perfect fit

Object Analytics is exactly the reverse of a relational database. All information about your business object is hooked into an object-centric data structure. This means that, for example, all information about each individual patient is available collectively in one “object instance”. What belongs together is stored together – and is available at your fingertips for holistic analyses.

With this object-centric representation, complex queries that analyze relationships between different sub-objects can be executed with high performance. Algorithms that were previously tedious to implement are now easy to apply. Novel algorithms – unimaginable before – become feasible (see Causal Discovery).

You may pull data from multiple sources and attach them as sub-objects (or recursive sub-objects), combining enterprise data from various origins into a single holistic picture. Object Analytics does not replace your relational databases. It just quickly aggregates data from potentially disparate sources into one object-centric view, allowing you to work with your business objects “as a whole”.

What that means for you

Typical sub-objects for your root business object are:

  • event streams such as the prescriptions, diagnoses, and procedures along the lifetime of the patient
  • manufacturing data with thousands of workpieces and the associated millions of process parameters, events, and messages along a production line

Easily answer questions pertaining to different sub-streams, e.g.

  • medications used according to a particular diagnosis
  • sequence of repair operations in relation to observed error messages

And ultimately, Object Analytics is the foundation for Causal Discovery. Feature engineering becomes obsolete – you don’t build features for a specific problem – you simply work with an object as a whole.

Object Analytics is moving from rows in a flat table to working with entire objects. Rethinking the world of statistics to deal with whole objects will push the field of machine learning and artificial intelligence into new dimensions.

What it looks like in practice

Some of our larger installations hold 60 million patients and 3 billion events in different sub-objects (in some projects up to 50 nested sub-objects) on a server with 128 cores. For each request, all cores will work in parallel, enabling us to serve numerous clients simultaneously.

The standard installation operates via a Docker container. Deploying our technology on a cloud platform such as AWS or just locally on your machine is extremely simple. The data scientist may use it as his local playground, and from there seamlessly turn proven solutions into a valuable enterprise deployment.