Cascade Imputation

This project contains the implementation of the "Cascade Imputation" described in our paper:

“Scalable Model-Based Cascaded Imputation of Missing Data”. Montiel, Jacob, Jesse Read, Albert Bifet, and Talel Abdessalem. In 2018 Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD), pp. 64-76. Springer, Cham, 2018.

Dependencies:

• scikit-learn
• pandas

---

Usage:

Required arguments:

• data: DataFrame to be processed D = (X,y)

• classifier: Classifier/regressor to be used for imputation. Options:

  "LR" - Logistic/Linear Regression

  "DT" - Decision Trees

  "RF" - Random Forest

  "ET" - Extreme Randomized Trees

• num_attr: List of numerical attributes

• cat_attr: List of categorical attributes

Optional arguments

• verbose: Enable verbose debug messages
• fp_log: Pointer to log file
• target_col: Name of the target column, "target" by default

Returns:

• DataFrame with imputed values D' = (X', y)