ImputeDB: Data Imputation as a Query Optimization

In order to study the placement of an imputation step, we create a logical imputation operator (along with respective physical instances) and incorporate its placement as part of the query plan optimization process in SimpleDB. We introduce measures for information loss and runtime for imputation operations, which outline the main trade-offs in the imputation placement. We add these measures into our cost estimation, allowing us to intelligently place the data imputation step during query planning. We show the trade-offs between efficiency and accuracy for simple data imputation models.