@InProceedings{Barbosa:2022:MaLeAp,
               author = "Barbosa, Ivan M{\'a}rcio",
          affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)}",
                title = "Machine learning applied to Amazonia-1 satellite power subsystem 
                         telemetry prediction",
                 year = "2022",
         organization = "International Astronautical Congress, 73.",
            publisher = "IAF",
             abstract = "This article presents the collection, exploratory data analysis, 
                         model training, evaluation, use of hyperparameters and 
                         implementation of the machine learning model that will be used to 
                         predict telemetry data from Amazonia-1 satellite developed by 
                         National Institute for Space Research (INPE). The payload data, 
                         which are acquired at the earth reception stations on INPE campus 
                         in CuiabA¡ (Mato Grosso State) and are processed, stored, and 
                         distributed, free of charge, are not part of the scope of this 
                         research work. AmazAnia-1 satellite is a satellite for remote 
                         sensing that was launched in 2021, uses the Multi-Mission Platform 
                         (MMP) as a service module and has an imaging camera named Wide 
                         Field Imager (WFI). It has 60 m of spatial resolution, 850 km 
                         width of the imaged strip and with 5 days revisit time. AmazAnia-1 
                         satellite has 715 telemetries with distinct data types (boolean, 
                         categorical, numerical) that will be used as dependent and 
                         independent variables. The amount of telemetry data generated 
                         daily is large and this makes manual analysis of this data 
                         unfeasible. Therefore, during the data preparation and 
                         manipulation phase, different attribute selection methods (e.g., 
                         filter, wrapper and embedded) are used. Then, the bagginng (e.g., 
                         Random Forest) and ensemble (e.g., XGBoost, AdaBoost) machine 
                         learning algorithms will be used to predict the values of the 
                         dependent variable of the telemetries of the electric power 
                         subsystem. of Amazon-1 satellite. For evaluation and performance 
                         of the machine learning model, the metrics Mean Absolute Error 
                         (MAE), Root Mean Square Error (RMSE) and (MAE) and R2 (coefficient 
                         of determination) will be used. At the end, the machine learning 
                         model with better quality and performance will be implemented by 
                         INPEs at TT\&C facilities.",
  conference-location = "Paris, France",
      conference-year = "18-22 Sept. 2022",
             language = "en",
           targetfile = "IAC-22,B6,1,9,x72024.brief.pdf",
        urlaccessdate = "05 jun. 2024"
}