Gradient Boosted Trees Predictor (Regression)
Has an R^2 of 0.705
Used every feature we had and was used as the best machine learning regressor based on its accuracy. There are several parameters that were chosen to be optimized, such as the number of models, from 100 to 400. Increasing the number of models, implies decreasing the learning rate a little, thus the learning rate became 0.05. The depth of the tree was also increased to 6. This optimization improved accuracy, from 0.705 to 0.735.
The model trained directly with the whole dataset has an even better accuracy of 0.842.
