Technology Focus: Application & Database Development

Audience: BI/Analytics Practitioner, Data Scientist

Level: 500

In this session I will introduce what is a mathematical model (i.e. a tool that describs with mathematical concepts how a system on which we want to act works) and a statistical model (which use more statistical tools). They both need an understanding (using physics, economics, etc. approaches) of what we want to act on.
Besides, they is Machine Learning which builds links between sections of a given system, without trying to explain them.
Machine Learning approach is really very efficient, essentially since it has no need of understanding, - so it has a kind universality - and since it compensate this lack by a heavy or long learning (selection, dependence building, etc.)
Yet, there is a approach, the Model-Data Coupling, that takes the best of both sides. It allows to incorporate what is taken for granted (by mathematical or statistical modeling) and to complete what cannot be understood (from Machine Learning).
I will explain on examples those approaches.

Prerequisites: No prerequisites. the talk will be illustrative, without mathematical or statistical techniques and bases on examples.