If you are translating from English into a Romance language, a useful starting point is the Grand dictionnaire terminologique maintained by Quebec’s Office de la langue francaise. The construction will be similar even if the exact words are different.
However, this only works because France has a significant information technology industry to provide the working terms, and because French-speakers in Quebec are strongly motivated to maintain their language. If your target language does not have equivalent properties, it may not be possible to create a native term with any staying power.
Your best approach might be to use the initials ML with a definition in parentheses after the first use. In English, we commonly use acronyms, such as FIFA, that have been created in other languages, but which have internationally accepted meanings.
Machine learning has been around since at least the sixties. It is possible that translations into your target language already exist, e.g. in course descriptions for computer science programs at the university level (and available on websites), or in popular articles reporting on advances in computer chess.
The word machine is used in machine learning as an adjective modifying the word learning, as others have pointed out. It’s a fairly common construction in computer science and engineering. Similar terms include automated, artificial, computational, etc., all of which have slightly different nuances and usages, and - importantly for your project - slightly different equivalents in other languages.