Online clustering approach

Question

Is there any "online" clustering approach? I mean that the procedure should be like this:

1. Can be fitted with the initial portion of data.
2. Can be updated with the upcoming batch of data.
3. The size of metadata to store should be significantly smaller than all available training data.

The data is GPS signals.

Answer 1

Yes, there are. An online variant of $k$-means clustering is pretty trivial and can be implemented by hand. For ready implementations of many different online algorithms, you can check Python's River package. Almost all of the online algorithms would store a learned representation of the solution rather than memorizing all the data, so would fit your third condition.