Model calibration with pyspark

This package provides a Betacal class which allows the user to fit/train the default beta calibration model on pyspark dataframes and predict calibrated scores

Setup

spark-calibration package is uploaded to PyPi and can be installed with this command:

pip install spark-calibration

Usage

Training

train_df should be a pyspark dataframe with score and label columns

from spark_calibration import Betacal
from spark_calibration import display_classification_calib_metrics
from spark_calibration import plot_calibration_curve


bc = Betacal(parameters="abm")


train_df = spark.read.parquet("s3://train/")

bc.fit(train_df) # training

print(bc.lr_model, a, b)

a,b -> coefficients of logistic regression model

lr_model -> pysparkML logistic regression model

Prediction

test_df is a pyspark dataframe with score as one of the columns. The predict function adds a new column prediction which has the calibrated score

test_df = spark.read.parquet("s3://test/")
test_df = bc.predict(test_df)

Pre & Post Calibration Classification Metrics

The test_df should have score, prediction & label columns. The display_classification_calib_metrics functions displays brier_score_loss, log_loss, area_under_PR_curve and area_under_ROC_curve

display_classification_calib_metrics(test_df)

Output

model brier score loss: 0.08072683729933376
calibrated model brier score loss: 0.01014015353257748
delta: -87.44%

model log loss: 0.3038106859864252
calibrated model log loss: 0.053275633947890755
delta: -82.46%

model aucpr: 0.03471287564672635
calibrated model aucpr: 0.03471240518472563
delta: -0.0%

model roc_auc: 0.7490639506966398
calibrated model roc_auc: 0.7490649764289607
delta: 0.0%

Plot the Calibration Curve

Computes true, predicted probabilites (pre & post calibration) using quantile binning strategy with 50 bins and plots the calibration curve

plot_calibration_curve(test_df)