search

kootenpv / whereami

Uses WiFi signals :signal_strength: and machine learning to predict where you are
GNU Affero General Public License v3.0
5.1k stars 247 forks source link

Added external data support #38

Closed AndrewDessin closed 7 years ago

AndrewDessin commented 7 years ago

I added in the ability to specify paths to both the model/trained data, and a path to find the current activity, or current X which we are trying to fit.

If a path is not specified, the normal flows continue.

coveralls commented 7 years ago

Coverage Status

Coverage decreased (-2.5%) to 88.177% when pulling 6842ae4223a61d762b3e55d47ef73a62a2cc9fb2 on AndrewDessin:master into 30befb04bc9b3107f4fd7bf3a39ccc879c0a34c9 on kootenpv:master.

kootenpv commented 7 years ago

Great! It works fine :) Just wondering: shouldn't learn also have extra arguments?

AndrewDessin commented 7 years ago

Yeah, I was a bit indecisive about it. The reason I didn't include external Learn args is because learn takes one location at a time, collects the data for that location, saves it to files, and calls the train_model() method. Instead I created a new train argument.

I assume that the external user will already have collected their data by locations and stored them as files by location. Therefore, I really only need to train the model. So that is why I added a new train argument, instead of adding arguments to learn. Do you agree?

kootenpv commented 7 years ago

Yea, that makes a lot of sense. Again, thanks! Nice PR :)