Clusterwise multiblock PLS

Ndèye Niang (1) , Stéphanie Bougeard (2, 3) , Gilbert Saporta (1)

1 CEDRIC - MSDMA - CEDRIC. Méthodes statistiques de data-mining et apprentissage
2 Laboratoire de Ploufragan - Plouzané
3 EPISABE - Epidémiologie, Santé et Bien-être

Clusterwise linear regression aims at partitioning a dataset into clusters characterized by their own regression coefficients. To deal with multiblock data, an extension of clusterwise regression to multiblock PLS is proposed. As this method is component-based, it may handle high dimensional data. The interest of the proposed method will be illustrated on the basis of a simulation study

Domaines: Statistiques [stat]

Ndèye Niang, Stéphanie Bougeard, Gilbert Saporta. Clusterwise multiblock PLS. SFC 2018, Sep 2018, Paris, France. ⟨hal-02471608⟩