Location-based species recommendation using co-occurrences and environment- GeoLifeCLEF 2018 challenge

Benjamin Deneu (1) , Maximilien Servajean (2, 3) , Christophe Botella (4) , Alexis Joly (1)

1 ZENITH - Scientific Data Management
2 ADVANSE - ADVanced Analytics for data SciencE
3 IBC - Institut de Biologie Computationnelle
4 BioSP - Biostatistique et Processus Spatiaux

This paper presents several approaches for plant predictions given their location in the context of the GeoLifeCLEF 2018 challenge. We have developed three kinds of prediction models, one convolutional neural network on environmental data (CNN), one neural network on co-occurrences data and two other models only based on the spatial occurrences of species (a closest-location classifier and a random forest fitted on the spatial coordinates). We also evaluated the combination of these models through two different late fusion methods (one based on predictive probabilities and the other one based on predictive ranks). Results show the effectiveness of the CNN which obtained the best prediction score of the whole GeoLifeCLEF challenge. The fusion of this model with the spatial ones only provides slight improvements suggesting that the CNN already captured most of the spatial information in addition to the environmental preferences of the plants.

Mots clés: Species; Plant

Domaines: Informatique [cs]; Sciences de l'environnement

Benjamin Deneu, Maximilien Servajean, Christophe Botella, Alexis Joly. Location-based species recommendation using co-occurrences and environment- GeoLifeCLEF 2018 challenge. CLEF 2018 - Conference and Labs of the Evaluation Forum, Sep 2018, Avignon, France. ⟨hal-01913241⟩