ESIROI · Maquettes Connexion
AccueilITS7 · UE7-DATA
ESI-SPI-CI-IN4-S7-UE5-EC3

SAE Data

FR EN ⬇ PDF
RéférentApply and evaluate machine learning algorithms and discuss their relevance for a given context.
ECTS2
CM / TD / TP4 / 0 / 2
Typesae

Viable
Viable100%
Complète86%
Manque pour « complète »
  • But du cours
  • Version EN relue

Acquis d'apprentissage visés

Process massive datasets using artificial intelligence techniques

Prérequis

  • Lecturers from DATA-related courses and engineering tools
  • Data analysis professionals

Programme

The company OcéanData Consulting, a data valorization consulting firm based in La Réunion, supports local businesses in their digital transformation. They entrust you with an analysis and modeling mission on a massive dataset from one of its clients — chosen from the following sectors: environment, tourism and mobility, public health, or energy and environment. The volume and nature of the data require the implementation of a suitable distributed architecture, from ingestion to result delivery.

  1. Big Data architecture design: select and justify the chosen technologies (distributed storage, computing framework, orchestration) in line with the constraints of volume, velocity, and variety of the data.
  1. Processing and modeling: implement one or more analysis methods from the following, depending on the project specifications:
  • Data mining and knowledge extraction (pattern mining, association rules)
  • Supervised classification (decision trees, SVM, random forests, neural networks)
  • Unsupervised classification / clustering (k-means, DBSCAN, hierarchical clustering)
  1. Delivery and valorization: present results clearly and in an exploitable format for a non-technical audience, justifying the methodological choices made.

Modalités d'évaluation

Skills will be assessed based on elements provided in the deliverable, which will be specified in advance to the engineering students.

Bibliographie

  • A documented Git repository containing all source code, analysis notebooks, and architecture deployment scripts.
  • A structured technical report including:
  • Description and justification of the chosen architecture (architecture diagram included)
  • Dataset description: origin, volume, quality, preprocessing steps
  • Applied analysis methodology, selected parameters, and their justification
  • Results obtained, their interpretation, and limitations
  • Compliance with best practices: code quality, reproducibility, version control
  • A results visualization dashboard (Jupyter Notebook, Streamlit, or equivalent)
  • A 20-minute oral presentation before a jury, including a live demonstration of the pipeline and results

Supports

All documents used during teaching sessions or resources mobilized. If needed, new documentary references will be provided.