Title: Feature Construction for Time Series and non-Time Series Datasets in Remote Sensing
Speaker: João E. Batista, LASIGE – DI/FCUL
When: May 6 (Thursday) at 12:00
Deforestation has serious implications on biodiversity, on rural communities that are dependent on forests for food and income, and on greenhouse gas emissions that are driving the global climate. One way to control deforestation is through the use of mechanisms that promote sustainable forestry and support economic development, like REDD+ of the United Nations. However, to be effective, these mechanisms require careful forest monitoring.
Having this into consideration, this work attempts to improve the quality of the results produced by the Machine Learning methods used by Remote Sensing experts for forest monitorization. To do so, this work promotes the use of automatic Feature Construction methods for multiclass classification in both Time Series and non-Time Series datasets that were extracted from satellite imagery. In particular, this work uses two methods based on Genetic Programming (Standard GP and M3GP) to perform Feature Construction and state-of-the-art methods (e.g., XGBoost and ROCKET) to classify the datasets.