Defining a cross_column for dependent categorical features

My features include two columns:

department

service

Because department and service are mostly "sparse" columns, they were defined as:

department

service

department_column = tf.feature_column.categorical_column_with_hash_bucket("department", 200, dtype=tf.int32)
feature_dict["department"] = tf.feature_column.indicator_column(department_column)

service_column = tf.feature_column.categorical_column_with_hash_bucket("service", 4000, dtype=tf.int32)
feature_dict["service"] = tf.feature_column.indicator_column(service_column)


But when we define this features as independent columns we're not describing the relation between this columns: service only makes sense only in the scope of particular department.

service

department

To describe the relation, it seems a good idea to use crossed columns.

The solution consist of two stages:

crossed_column

crossed_column

embedding_column

indicator_column

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.