Selecting regression algorithms

By default, ML.NET supports the following regression model trainers:

• FastTree: A decision tree algorithm, which we discussed earlier in this chapter. Decision trees are fast, simple, and easy to understand, and they are not sensitive to scaling. However, if you allow them a high number of leaves, they may overfit the training data at the expense of accuracy on data they’ve not seen.
• FastTreeTweedie: Almost identical to FastTree but built for a specific distribution of data, referred to as a Tweedie distribution. See the Further reading section for more details.
• LightGbm: Uses gradient boosting and Microsoft’s LightGbm framework for high-performance decision trees.
• FastForest: A random forest algorithm. This approach uses a voting ensemble of decision trees, with each tree’s prediction constituting part of the overall “vote” for the final prediction.
• LbfgsPoissonRegression: A linear algebra approach...