search

PokemonGoers / PredictPokemon-2

In this project we will apply machine learning to establish the TLN (Time, Location and Name - that is where pokemons will appear, at what date and time, and which Pokemon will it be) prediction in Pokemon Go.
Apache License 2.0
9 stars 3 forks source link

Gym/Pokestop feature #45

Closed MatthiasBaur closed 7 years ago

MatthiasBaur commented 8 years ago

Retrieve the information from the big DataDump and create a corresponding feature.

MatthiasBaur commented 8 years ago

This issue is currently worked on in the Gym_data branch. See https://github.com/PokemonGoers/PredictPokemon-2/tree/Gym_data

bensLine commented 7 years ago

@goldbergtatyana fyi i created those plots when i worked on the gym and pokestop feature, might be interesting for you. Most sightings in the 600k set strongly correlate with the locations of the pokestops and gyms as expected. However, there are a few outliers. Also the distribution of pokemons, gyms and pokestops varies a lot. if you look at the Europe images there is always much more data for the Czech Republic. Maybe the data drop just provides more sightings there but it seems a bit odd. Probably there are much more players... gym locations europe gym locations pokemon locations - data dump pokemon locations europe - data dump pokestop locations europe pokestop locations

goldbergtatyana commented 7 years ago

very interesting plots @bensLine that are much appreciated! We should show them in our facts and statistics site if we will have one and in the presentations if we will give any :))

Your result shows that pokestops (are they really included in the file?) and gyms are highly important for the prediction of pokemons (which is somewhat we figured out on our own :-))! The fact that there are more gyms in Czech replublic is good for the players living there, as they will get most precise results. Hoperfully, they will like us and tell the whole world how great we are!!