deployment - h2o pojo on test data with extra columns than the model trained on and sometimes missing columns from the train dataset

Question

I have created my model POJO, I have to keep my columns in same order with same datatype when generating predictions using Hive UDF? what is the cleanest way to ignore extra columns and add the columns which are present in train data set but not in test data set, my all columns are either double or long.

标签： deploymentpojoh2o

Answer 1

如果您使用 Easy 包装器，它会自动为您执行此操作。

如果您不使用 Easy 包装器，那么您需要发明相同的行为。

使用 Easy 包装器，新列将被忽略，缺失的列将被视为 N/A。