metadata

language:
  - en
thumbnail: null
tags:
  - classification
  - gradient boosted trees
  - keras
  - TensorFlow
library_name: generic
libraries: TensorBoard
license: apache-2.0
metrics:
  - accuracy
model-index:
  - name: TF_Decision_Trees
    results:
      - task:
          type: structured-data-classification
        dataset:
          type: census
          name: Census-Income Data Set
        metrics:
          - type: accuracy
            value: 96.57
          - type: validation loss
            value: 0.227394

TensorFlow Decision Forests for structured data classification

Use TF's Gradient Boosted Trees model in binary classification of structured data

Build a decision forests model by specifying the input feature usage.
Implement a custom Binary Target encoder as a Keras Preprocessing layer to encode the categorical features with respect to their target value co-occurrences, and then use the encoded features to build a decision forests model.

The model is implemented using Tensorflow 7.0 or higher. The US Census Income Dataset containing approximately 300k instances with 41 numerical and categorical variables was used to train it. This is a binary classification problem to determine whether a person makes over 50k a year.

Author: Khalid Salama
Adapted implementation: Tannia Dubon