Features#

This page lists all the features of YDF.

Learning algorithms#

  • CART

  • Random Forest

  • Gradient Boosted (Decision) Trees

  • Distributed Gradient Boosted (Decision) Trees

Meta-learning algorithm#

  • Automatic hyper-parameter optimizer

Supported of problems#

  • Classification (binary and multi-class)

  • Regression

  • Ranking

  • Uplift

  • Weighted

Supported input features#

  • Automatic feature type detection and dictionary building.

  • Numerical

  • Categorical

  • Boolean

  • Categorical-set

  • Missing

Inference#

  • VPred

  • QuickScorer Extended

  • Get leafs

Model evaluation#

  • Classification

    • Accuracy

    • AUC (Area under the curve) of the ROC curve

    • AUC of the Precision-Recall curve

    • ROC curve

    • Precision-Recall curve

    • Precision @ Recall

    • Recall @ Precision

    • Precision @ Volum

    • Recall @ False positive rate

    • False positive rate @ Recall

    • Cross-entropy loss

  • Regression

    • RMSE

    • MSE

  • Ranking

    • NDCG

    • MRR

    • Precision @ 1

  • Uplift

    • AUUC

    • Qini

    • Cate Calibration

  • Confidence interval

    • Bootstrapping

    • Closed-form (for a subset of metrics)

API#

  • CLI

  • C++

  • Python / TensorFlow

  • Go

  • JavaScript

Dataset format#

  • CSV

  • TF.Example proto of in TFRecord container

Model analysis#

  • Variable importance

  • Decision tree plotting

  • Tree structure statistics