Data preparation.
The logs were collected from the gateway service through which the Watson APIs are accessed, and contain 6.3M entries.
Each entry contains the requested URL, the HTTP method, and the response status code.
We filtered log entries by their HTTP method, keeping the ones with GET, POST, DELETE, and PUT, and removing those with OPTIONS and HEAD, which do not immediately correspond to resource manipulations.
Other methods were not present in the logs.
We also preserved only log entries where the reported HTTP response status code is in the 200-299 range, denoting that the request succeeded.
Applying these filters reduced the number of log entries to 5.9M.