Pokemon tracker in selected places using inferential statistic
Recently one phone game spread through the whole world, andΒ cause a lot of interesting topic about the technique behind the game. One of the hot topic is: How to predict the pokemon spawning position? Β
HereΒ I developed an app to predict the pokemon spawning position and the probability of spawning at that position. This is the exciting news for the fans of the game, who can easily find the possible spawning position and collect the rare desired pokemon. Next let me introduce my app.
- Features Selection. Β Β Β Β Β Β the data was downloaded from kaggleΒ with 2906021 observation and 208 variables. 7 variables was selected, including the geography coordinates(longitude and latitude), the pokemon Id, name,appeared Time Of Day and the city.
- MethodologyΒ . Β Β Β Β Β Β Β Β Β Β group the data by locationΒ (longitude and latitude), and count the population of the specific pokemon appeared at thatΒ location, the circle on the map show the appeared locationΒ in history, the intensity of the circle's color and the radius of the circles proportional to the populations. So the darker and larger circle indicate higher possibility that the pokemon appear at that position.
- Function of the app. Β Β Β Β In the application(see figures below), the customer can:
- select the cityΒ Β Β Β Β Totally 98 citiesΒ of different country and different continent areΒ avaliable, and Β more and more cities are going to added.
- select the target pokemon. Β Β Β Β Β Β All 143 kinds of pokemon were collectedΒ and the statistic data of the pokemon is shown in the table above the map.
- select the type of street map. Β Β Β Β There are 5 kinds of map to chooseΒ "HERE.hybridDay", "Stamen.TonerLines", "Stamen.Terrain", "CartoDB.Positron", "Esri.WorldImagery".
- select the zoom level Β Β Β Β Β Β Β Β Β There are 18 levels of zoom can be choose. Β Β The highΒ level can be chose to show the detail of the destination, and the low to show the large area of the location and more spawn location to choose.
Figure 1 show the pokemon (Pidgey) spawn distribution and frequent at each spawnΒ location
Figure 2 select the city: Los Angels and adjust the zoom level to 16, and thisΒ show very small range and explicit spawn location
Figure 3, using theΒ the satellite map "Esri.WorldImagery" to search the very rare kind of pokemon(Pikachu) and investigate the surrounding of the location
Figure 4Β the zoomed central park map showing Β the explicit pokemon spawn locationΒ and the traffic.
Future Step:
Apply more features and machine learning , get more accurate prediction
Actually, there is one very complicated equation to generate the pokemon spawns. In the future, I am going to add more features, such as wind speed, temperature, resident population density, the local time, the moving speed, and apply the machine learning algorithm Β to track the pokemon more accurate.