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

Feature/population #34

Closed marwage closed 8 years ago

marwage commented 8 years ago

The script gives you for a given latitude and longitude the city/town/village with nominatim.openstreetmap.org and for that place the population based on WolframAlpha Would be nice if you look over the code ;)

28

bensLine commented 8 years ago

Cool, just tested the script! Unfortunately, it takes veeery long to get the population. the call to openstreet map for the city name is way faster.

If we want to use the data the script needs to be implemented as feature source. However, we are mainly interested in the population, but if it takes that long I'm not sure if that feature will be applicable :/ Seems like wolframalpha is here the problem as the requests take a very long time. I guess we need to find another source if we cannot speed up the requests to wolframalpha.

goldbergtatyana commented 8 years ago

Please ask the https://github.com/PokemonGoers/PokeData group if the ran into the same issue and how they resolved it.

bensLine commented 8 years ago

@goldbergtatyana I googled a little bit and the problem seems to be a general one of wolfram alpha. at least for the free version the response time is very slow - also on their website e.g. https://www.wolframalpha.com/input/?i=munich

goldbergtatyana commented 8 years ago

@sacdallago @gyachdav do you know if we have an elegant solution for this issue?

gyachdav commented 8 years ago

hmm if wolfram alpha is slow then we will indeed need a different source. How's Wikipedia in thatvregard?

sacdallago commented 8 years ago

I like @gyachdav 's idea.. If we know the city name, we might just check with wiki the population density.

Otherwise: how does wolfram alpha actually do it? Is there someway we can completely mine their databases? Would it be possible to get some YxY patches with population density and then just go by nearest neighbor?