Reporter Topic:Use of Machine learning to predict Species distribution models under climate change scenarios
Reporter:LEK Sovan Director，University of Toulouse (France)
Location:Academic Hall of State Key Laboratory
LEK Sovan is Emeritus Professor at University of Toulouse (France). His research is mainly in Fish Community Ecology and Ecological Modelling. He participated to several EU projects and coordinated several of them in the fields of Ecology and Global changes. He participated also to several bilateral projects with several Asian countries, like China, Thailand, Vietnam, Indonesia, Cambodia and Korea. He also participates to several project for the students mobility e.g. Marie-Curie Keybioeffects or Erasmus Mundus TECHNO & LOTUS. He contributes to several Erasmus+ Capacity Building projects: UNICAM, NutriSEA, CONSEA.He published 3 books and more than 250 papers in the SCI journals. His h-index = 36. The average of his citation is around 500 per year. He is editorial member of Ecological Modelling and Ecological Informatics.
Machine learning is a branch of artificial intelligence that employs a variety of statistical, probabilistic and optimization techniques that allows computers to “learn” from the examples and to detect characteristic patterns from large, noisy or complex data sets. This capacity is particularly well suited to ecological modelling, especially those that depend on number of variables measured and their often nonlinear relationships.
As a result, machine learning is frequently used in population and ecosystem studies. More recently machine learning has been applied to pattern the community structure and prediction. This latter approach is particularly interesting as it is part of a growing trend towards personalized, predictive ecology. In assembling this review we conducted a broad survey of the different types of machine learning methods being used, the types of data being integrated and the performance of these methods in ecological prediction and diagnosis. A number of trends are noted, including a “older” technologies such artificial neural networks (ANNs) instead of more recently developed or more easily interpretable machine learning methods such as tree family models, Support Vector Machine and etc. To test the predictive power, large spatio-temporal fish data from different spatial scales were used. Results showed that Machine learning can predict well the distribution of species over space as well as over time.