[WEEK 5 — Wi-Fi Based Indoor Positioning]
Team Members: Burak Emre Ozer, Huzeyfe Kocabas
This week we are going to restrict our data set. So we are just going to predict the position according to only one building. We look forward to the results of this impact. Let’s begin…
First, we remove the data with latitude and longitude values of zero. Besides, only to select the TI building set and filtered the buildingID to 0.
Then we train random forest regressor model and got the average error value: 8.30 meters as seen from the side location of each points on the map.
The error value we found in the past weeks was almost halved. The results are promising.