Main data mining prototypes

SITS-P2miner : Pattern-Based Mining of Satellite Image Time Series

SITS-P2miner is a mining system for extracting patterns from Satellite Image Time Series. It includes four main modules for pre-processing, pattern extraction, pattern ranking and pattern visualization. It is based on the extraction of grouped frequent sequential patterns and on swap randomization. (download zip archive from the SITS-P2miner home page)

Developed in collaboration with the LISTIC Lab. (contact: N. Meger).

DMT4SP : extraction of episodes and of sequential patterns under constraints

Click here for the home page of the dmt4sp tool.

EvoMove : an evolutionary musical companion for dancers

A system using IMU sensors and an evolutionary subspace clustering algorithm to analyze dancer motions and create live soundtrack. Developed in collaboration with S. Peignier and other members of the Beagle team (Guillaume Beslon, Jonas Abernot, Anthony Rossi and Leo Lefebvre) within the EvoEvo european project.

DEMO: Click here to see a working session.

Other selected prototypes

Here are mentioned only the main data mining prototypes, developed under contract with a company, or distributed and used by other academic teams in other labs.

For each, my participation covers several of the following aspects: pattern specification / pattern ranking measure / algorithm design / proofs of soundness and completeness / implementation / parameter tuning methods / pre-processing and post-processing techniques.

  • Marguerite (PhD Ieva Mitasiunaite) Differential mining of string sets under constraints.
  • WinMiner (PhD Nicolas Meger) Finding episode rules and their maximal confidence temporal windows in sequences of events.
  • Go-Spec (PhD Marion Leleu) Extracting sequential patterns in sequences containing contiguous event repetitions.
  • ACE3 (PhD Thomas Daurel) Extracting classification rules with delta-free antecedents.
  • ACMiner (PhD Artur Bykowski) Finding delta-free sets and delta-strong rules in Boolean transaction databases.

Special thanks to users, experts in application domains, and other data mining researchers that contributed to these works.