optimize_station.py.txt

import matplotlib.pyplot as plt
import geopandas as gpd
from shapely.geometry import Point
import matplotlib.pyplot as plt
import pandas as pd
from pandas import DataFrame
from scipy.cluster.vq import kmeans2
import numpy as np
from scipy.optimize import fmin,basinhopping,rosen, differential_evolution
import math
import scipy.optimize as optimize

crimes = gpd.read_file("Police_Incidents_2015.geojson")
precinct = gpd.read_file("precincts.geojson")
stations = gpd.read_file("Police_Stations.geojson")

# Define a Function to make clusters in the data based on the feature vector and the no. of clusters required.
def clustering (data,num_clusters):
    x, y = kmeans2(data, num_clusters , iter = 20)
    return (x,y)

# Define a function to make a plot of the crime location and the centeroids based on clusters.   
def plot_clusters(data,centeroids,clusters):
    plt.scatter(data['Latitude'], data['Longitude'], c=clusters, alpha=0.33333)
    plt.scatter( centeroids[:,0],centeroids[:,1],c='red',marker= 'x',alpha=1 )
    plt.show()
    
# Read the CSV file into a DataFrame
crime = pd.read_csv('Police_Incidents_2015.csv')
pd.set_option('precision',8)
# Create a dataframe with only the required features for clustering
crime_location = DataFrame()
crime_location['Latitude'] = crime['Lat']
crime_location['Longitude'] = crime['Long']
# Specify the no. of clusters required
num_clusters = 13

# Function call to create clusters using Kmeans2

centeroid,cluster =clustering(crime_location,num_clusters)
# Function to plot the Co-Ordinates on the map as per clusters.
#plot_clusters(crime_location,centeroid,cluster)
array2=np.array([[45.00812948,-93.2468554],
[44.94817514,-93.23445848],
[44.9913555,-93.30508653],
[44.94648138,-93.27755368],
[44.98079656,-93.27215832],
[44.96877199,-93.24745199],
[44.96939228,-93.26355705],
[44.96763002,-93.27942831],
[44.98815103,-93.25670347],
[45.01235066,-93.29454588],
[44.96407055,-93.27764843],
[44.94912052,-93.29810577]])
crime_centers_cluster=pd.DataFrame(centeroid,columns=["Latitude","Longitude"])
police_stations=pd.DataFrame(array2,columns=["Latitude","Longitude"])
police_stations["distances"]=0
total_distance=pd.Series(np.zeros(len(centeroid)))

def dist1(x):
    global police_stationsr
    global centeroid
    global total_distance
    temp=pd.DataFrame([[x[0],x[1],0]],columns=["Latitude","Longitude","distances"],index=[12])
    police_stations1=police_stations.append(temp)
    total_distance=pd.Series(np.zeros(len(centeroid)))
    for i in range(len(centeroid)):
        police_stations1["distances"]=np.sqrt((police_stations1["Latitude"]-crime_centers_cluster.ix[i]["Latitude"])**2+(police_stations1["Longitude"]-crime_centers_cluster.ix[i]["Longitude"])**2)
        total_distance[i]=min(police_stations1["distances"])
    #print police_stations1
    return sum(total_distance)*1000000
#bounds contains the boundaries for both latitude and longitude
bounds=[(44,46),(-95,-92)]     
ret=differential_evolution(dist1, bounds,tol =.001,maxiter =10**3,mutation=(0.00001,0.00001),polish =True,popsize =1000)  
print ret
print "minimum distance {}".format(ret.fun/1000000)
new_police_station_latitude=ret.x[0]
new_police_station_longitude=ret.x[1]

crime_centers_cluster["geometry"] = zip(crime_centers_cluster.Longitude, crime_centers_cluster.Latitude)
crime_centers_cluster["geometry"] = crime_centers_cluster["geometry"].apply(Point)

new_station = gpd.GeoSeries([Point(list(reversed(ret.x)))])
centroids = gpd.GeoSeries(crime_centers_cluster["geometry"])

ax=precinct.plot()
ax = crimes.plot(ax=ax)
ax=stations.plot(ax=ax, marker="x", color="blue", mew =10, markersize=20)
ax = centroids.plot(ax=ax, marker="x", color="red", mew=10, markersize=20)
new_station.plot(ax=ax, marker="x", color="green", mew=10, markersize=20)
plt.title("Crimes in Minneapolis - 2015")
plt.show()