Label Encoding

• also called Integer Encoding
• Each unique category is assigned an integer value
• e.g. “red” → 0, “blue” → 1, …
• easy reversible
• can only be used when a ordinal relationship between the labels exist, e.g. winner ranking in string (“first”, “second”, “third”)
  • if not and still used, can result in poor performance and unexpected results
• numeric representations have a natural ordered relationship between each other and the models are able to understand that relationship