Sofware(s) :


enneade.gifKhiops - Ennéade : The tool realizes 'something' close to a "supervised kmean". The papers which describe parts of the process incorporated in Ennéade are (up to now) indicated below.

This software is used by business units inside Orange. If you are interested by the idea of "supervised clustering" : this sofware is distributed inside and outside Orange.

  • [2015] "Classification à base de clustering ou comment décrire et prédire simultanément", Oumaima Alaoui Ismaili, Vincent Lemaire, Antoine Cornuéjols to appear in "Rencontres des Jeunes Chercheurs en Intelligence Artificielle (RJCIA)", Rennes, 2015
  • [2015] "An Initialization Scheme for Supervized K-means" Vincent Lemaire, Oumaima Alaoui Ismaili, Antoine Cornuéjols, in International Joint Conference on Neural Networks (IJCNN), IEEE, Ireland, 2015
  • [2015] "Supervised pre-processings are useful for supervised clustering", Oumaima Alaoui Ismaili, Vincent Lemaire, Antoine Cornuejols, in the Springer Series “Studies in Classification, Data Analysis, and Knowledge Organization”
  • [2014] "A supervised methodology to measure the variables contribution to a clustering", Oumaima Alaoui Ismaili, Vincent Lemaire, and Antoine Cornuéjols - in International Conference on Neural Information Processing (ICONIP), Kuching, Sarawak, Malaysia from 3 - 6 November 2014
  • [2014] "K-means clustering on a classifier-induced representation space : application to customer contact personalization", in Annals of Information Systems, Springer - Special Issue on Real-World Data Mining Applications

  • Software Registration (with Carine Hue, Nicolas Voisine and Françoise Fessant): IDDN   : FR.001.520021.000.S.P.2012.000.00000 to the Agency of Software Protection.



    khiops_interpretation.gifKhiops - Interpretation [or named "Kawab" in scientific publications] :
    This software is used by business units inside Orange. If you are interested by the idea of "individual intepretation" : this sofware is distributed inside an doutside Orange.

    Description : Khiops Interpretation realizes the interpretation of the scores of a naïve Bayesian predictor for classification problems. The predictor to interpret must be described by a dictionary produced by the Orange scoring tool Khiops (version V8.0 (only mono-table classifier)). The tool is available both in user interface and in batch mode. This guide details the tool inputs and outputs and the interpretation actions. This tool could be useful for business units (for example) or research centers who want to analyze classification results with input values exploration.

    Features : The tool delivers two aspects of the interpretation

  • A “Why” analysis: the importance or contribution of each input variable for each instance for a given value of the target class. Several contribution indicators are available, among them indicators of the state of the art.
  • A “How” analysis: it gives the actions to realize in order to reinforce the predicted probability of a given class. This analysis can be very useful in commercial campaigns.

  • An illustration is available here : "Correlation Analysis in Classifiers", Vincent Lemaire, Carine Hue, Olivier Bernier, in “Handbook of Research on Data Mining in Public and Private Sectors: Organizational and Government Applications

    Publications : The tool has been the subject of a few publications (sometimes in French, sometimes in English and therefore redundant):

  • [2012] "A Complete Data Mining process to Manage the QoS of ADSL Services", F. Fessant & V. Lemaire in the workshop WAITS 2012 (Workshop on Artificial Intelligence for Telecommunications & Sensor Networks held at ECAI 2012, Montpellier)
  • [2011] "Gestion de la QoS des services ADSL à l’aide d’un processus de data mining", Vincent Lemaire and Françoise Fessant, workshop "Data Mining, Applications, Use Cases and Success Stories" joint to the conference 'Extraction et Gestion des Connaissances (EGC), Brest, 2011.
  • [2010] Exhibition : "KAWAB : Un outil pour explorer les corrélations existantes dans un classifieur naïf de Bayes", Vincent Lemaire, in RFIA (Reconnaissance des Formes et Intelligence Artificielle), Caen, Fevrier 2010
  • [2010] "Correlation Analysis in Classifiers", Vincent Lemaire, Carine Hue, Olivier Bernier, in "Handbook of Research on Data Mining in Public and Private Sectors: Organizational and Government Applications"
  • [2009] "Correlation Explorations in a Classification Model", Vincent Lemaire, Carine Hue, Olivier Bernier in the workshop Data Mining Case Studies and Practice Prize, SIGKDD 2009
  • [2009] "Exploration des corrélations dans un classifieur, Application au placement d’offres commerciales", Vincent Lemaire, Carine Hue, in Extraction et Gestion des Connaissances (EGC), Strabourg.

  • Note, we did this work [in 2010-2011] but :

  • For those who know the LIME method.. You may think about the “Why” analysis of the tool as being close to the recent (known) method LIME of M. Tulio Ribeiro [...] (2016).
  • For those who know “Counterfactual Explanations”. You may think about the “How” analysis of the tool as being close to the recent work of S. Watcher [...] (2017).

  • Software Registration (with Carine Hue): IDDN   : FR.001.170013.000.S.P.2012.000.31500 to the Agency of Software Protection.