How to obtain value from GDBT Regressor?

General Discussion
#1

Hello, I have Regressor code from example

# Build the input function.
train_input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
    x={'x': x_train}, y=y_train,
    batch_size=batch_size, num_epochs=None, shuffle=True)
test_input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
    x={'x': x_test}, y=y_test,
    batch_size=batch_size, num_epochs=1, shuffle=False)
# GBDT Models from TF Estimator requires 'feature_column' data format.
feature_columns = [tf.feature_column.numeric_column(key='x', shape=(num_features,))]

gbdt_regressor = tf.estimator.BoostedTreesRegressor(
    n_batches_per_layer=num_batches_per_layer,
    feature_columns=feature_columns, 
    learning_rate=learning_rate, 
    n_trees=num_trees,
    max_depth=max_depth,
    l1_regularization=l1_regul, 
    l2_regularization=l2_regul
)

gbdt_regressor.train(train_input_fn, max_steps=max_steps)

gbdt_regressor.evaluate(test_input_fn)

Can someone advise me how to use the function to obtain value estimate from parameters? Whole example is here: TensorFlow-Examples/gradient_boosted_trees.ipynb at master · aymericdamien/TensorFlow-Examples · GitHub
Thank you

#3

@Jan_Benes_III Welcome to the Tensorflow Forum !

In a Gradient Boosted Decision Tree (GBDT) model, you can obtain value estimates from the parameters by evaluating the model on new data points. The exact process can vary depending on the library or framework you are using for GBDT modeling. I’ll provide a general outline of the steps involved in obtaining value estimates from GBDT parameters:

  1. Train the GBDT Model: First, you need to train a GBDT model on your training data. You can use popular libraries like XGBoost, LightGBM, or scikit-learn for this purpose. The training process will involve learning the optimal parameters for the decision trees in the ensemble.
  2. Load or Reconstruct the Trained Model: After training, you’ll either have a trained model saved to a file or an object in memory, depending on the library you’re using. You will need to load or reconstruct this trained model for inference.
  3. Prepare New Data: To obtain value estimates, you’ll need new data points that you want to make predictions for. This new data should have the same features as the training data used to train the GBDT model.
  4. Predict Values: Use the loaded or reconstructed GBDT model to make predictions on the new data. The specific method for making predictions varies depending on the library or framework you’re using. In general, you’ll call a prediction function or method on the model object, providing the new data as input.

Let us know if this helps!