M- DATA SCIENCE

Technological development is accompanied by the creation of data in numerous activity sectors: measurements collected by sensors, (for example on production lines, for the analysis and management of pollution), results of numerical modelling or simulation, consumption/production of energy, meteorological data, traces left on internet, etc. Data Science (or Data Analytics) aims to exploit this data. One of the branches of Data Science is Big Data, which concerns extreme cases of vast Volumes of Varied format data arriving at Very high speed. We cannot process such data very easily today so the question can be viewed as a multi-disciplinary challenge, and is thus the subject of the Big Data interdisciplinary specialisation. In terms of teaching, Data Science as a whole is based on statistical and algorithmic methods associated with clearly identified fields: statistical learning, machine learning and data mining. These are the methods the Data Science major seeks to develop.Students will be provided with an overview of the main classes of learning methods, supervised and non-supervised. The training will also give practical courses on data analysis thanks to numerous case studies, from varied contexts.

The major is composed of 4 units. The first two units provide the fundamentals necessary for processing all types of data. The final two units will deal with two particular important cases: temporal and spatial data (Unit 3) and costly computer experiments (Unit 4).

Unit 1. Probabilistic basis
This introductory module provides the techniques to describe, visualise and understand data. It also gives the tools for the next units.

Unit 2. Statistical learning
Probability models and algorithmic methods for data learning.
Application to a professional problem, coupled with Unit 3.

Unit 3. Machine learning
Statistical methods and probability methods for the analysis of data correlated in time.
Application to a professional problem, coupled with Unit 2.

Unit 4. Metamodeling and Optimization
A field in its own right (key words: computer experiments, meta-modelling) aiming to exploit data which is costly to obtain (simulation durations too long) using simplified models obtained by function approximation methods.

Listed below are some examples of engineering programmes combing with another major or specialisation, using Data Science skills.

Comments

(1) This profile corresponds to big data exploitation, in particular internet data.

(2) The use of statistical methods is widespread in the healthcare field, for example, predicting the future reaction of a patient to a course of medical treatment, based on the data from preliminary blood tests, to detect variables influencing reactions.

(3) Data science tools are used for example to forecast electricity consumption/production from meteorological data, handy for adjusting production levels but also for feasibility/profitability studies for future wind turbine parks (or solar panels).Feasibility questions are also raised for the use of sensors in private habitations. Finally numerical simulations are widely used in the nuclear energy sector for risk and hazard studies. 

(4) A profile sought by the automobile and aeronautical industries. For example the mathematical processing of numerical simulation results is used to lighten vehicle weight, an obsession with car manufacturers, while maintaining adequate safety levels. These methods are also used to produce less polluting engines, etc.

(5) The microelectronics industry is brimming over with data which can be exploited for process control and quality improvement

(6) Data on car driving speed for example is collected from mobile telephones so as to improve traffic flow.