MACHINE EARTHING
VAN VEEN
THE PROBLEM
Through much effort, dedication, and specific knowledge, experts in many different fields interpret satellite data daily to make decisions that benefit our society. However, due to the enormous amount of types, sources, and variables in this data, judging how beneficial an action is can be extremely complex. That’s due to many factors that influence this analysis may not be considered in a human evaluation.
THE SOLUTION
Decisions made based on satellite data affect our lives daily, from farming to weather forecasting, but how can we tell how beneficial these decisions are? We propose a machine learning algorithm that will judge how beneficial they can be. It will be able to learn which factors are most relevant for each situation. After assisted training, the algorithm will evaluate the decision on a scale from 1 to 5, and based on previous data will define how beneficial the action was. The importance remains on the self-learning factor, which after training, allows it to establish relationships between factors that are not always considered in the human evaluation due to their complexity.
LEARN MORE
Through much effort, dedication, and specific knowledge, experts in many different fields interpret satellite data daily to make decisions that benefit our society. However, due to the enormous amount of types, sources, and variables in this data, judging how beneficial an action is can be extremely complex. That’s due to many factors that influence this analysis may not be considered in a human evaluation. A machine, meanwhile, after a supervised training process and many tests, may find a connection between the relative humidity of the air and combating deforestation, for example.
We propose an algorithm that will be trained to have an accurate perception of which distinct factors characterize a successful decision, regardless of which category belongs to or which variables are included in the process. To accomplish this, it will need to be submitted to the training process, which consists of having access to the database used to make the decision and a score given by professionals for the same action. During the training phase, the algorithm will identify which factors are most relevant for a successful action. Once will be capable of measuring how beneficial a decision was without the use of external sources. We intend to use ‘’Python’’ combined with libraries like “Pytorch” to build the algorithm.
TEAM MEMBERS
Arthur Mancio Sales
Developer
Guilherme Martiniano Rodrigues
Developer
Maria Eduarda Alves Vilas Boas
Researcher & Translator
Yasmin Daher Macri
Designer
