Use S2 library to create cell id features from lat/lon

bensLine commented 8 years ago

S2 JS library: https://github.com/Daplie/s2-geometry.js
nice blog post about S2 library: http://blog.christianperone.com/2015/08/googles-s2-geometry-on-the-sphere-cells-and-hilbert-curve/

bensLine commented 8 years ago

I tested the library and approximated the cell distances for the different levels. I'll add the code when https://github.com/PokemonGoers/PredictPokemon-2/issues/4 is merged, as it is only a simple script now..

level 0, avg distance: 10007.543398km 10007543.398m
level 1, avg distance: 6227.188138km 6227188.138m
level 2, avg distance: 3219.191285km 3219191.285m
level 3, avg distance: 1501.842929km 1501842.929m
level 4, avg distance: 724.902798km 724902.798m
level 5, avg distance: 371.812270km 371812.270m
level 6, avg distance: 188.665493km 188665.493m
level 7, avg distance: 94.034835km 94034.835m
level 8, avg distance: 46.852384km 46852.384m
level 9, avg distance: 23.406853km 23406.853m
level 10, avg distance: 11.698556km 11698.556m
level 11, avg distance: 5.851886km 5851.886m
level 12, avg distance: 2.925292km 2925.292m
level 13, avg distance: 1.462483km 1462.483m
level 14, avg distance: 0.731261km 731.261m
level 15, avg distance: 0.365620km 365.620m
level 16, avg distance: 0.182811km 182.811m
level 17, avg distance: 0.091405km 91.405m
level 18, avg distance: 0.045702km 45.702m
level 19, avg distance: 0.022851km 22.851m
level 20, avg distance: 0.011426km 11.426m
level 21, avg distance: 0.005713km 5.713m
level 22, avg distance: 0.002856km 2.856m
level 23, avg distance: 0.001428km 1.428m
level 24, avg distance: 0.000714km 0.714m
level 25, avg distance: 0.000357km 0.357m
level 26, avg distance: 0.000179km 0.179m
level 27, avg distance: 0.000089km 0.089m
level 28, avg distance: 0.000045km 0.045m
level 29, avg distance: 0.000022km 0.022m
level 30, avg distance: 0.000011km 0.011m

goldbergtatyana commented 8 years ago

Nice job @bensLine! Does the library allow to exclude regions such as seas, oceans and Antarctica? Would be nice to work with only those points that are really reachable by users.

bensLine commented 8 years ago

relates to #25. S2 cell id features have been implemented as example feature source. the script to print the average distance is also included.

bensLine commented 8 years ago

@goldbergtatyana, unfortunately, we do not get any further information from the S2 library, it only provides the cell id. if we extract terrain features #7 we can infer this information.

PokemonGoers / PredictPokemon-2

Use S2 library to create cell id features from lat/lon #5