def getorigin(ilat, ilong, popemp, mbta, zipscale, subwayscale):
"""
Compute features for where people live, work, and ride the subway.
"""
# locations of where people live
latbyzip = popemp['latitude'].values
longbyzip = popemp['longitude'].values
# locations of where people work
popbyzip = popemp['HD01'].values
workbyzip = popemp['EMP'].values
# replace nans with average value
finite = workbyzip * 0 == 0
mwork = workbyzip[finite].mean()
nans = workbyzip * 0 != 0
workbyzip[nans] = mwork
# locations of where people ride the subway
latbysubway = mbta['latitude']
longbysubway = mbta['longitude']
subwayrides = mbta['ridesperday'].values
# get the distance from given location to all zip codes
distancevec = densitymetric.distvec(latbyzip, longbyzip, ilat, ilong)
# compute the coupling factor to all zip codes
couplingzip = densitymetric.zipcouple(distancevec, zipscale)
# use the weighted sum as the origin score
# normalize by number of zip codes
nzip = len(workbyzip)
originpop = np.sum(popbyzip * couplingzip) / nzip
originwork = np.sum(workbyzip * couplingzip) / nzip
# get the distance from given location to all subway stops
distancevecsubway = densitymetric.distvec(latbysubway, longbysubway,
ilat, ilong)
# coupling efficiency between this station and all subway stops
couplingsubway = densitymetric.zipcouple(distancevecsubway, subwayscale)
# use the weighted sum as the origin score
# normalize by number of subway stops
nsubway = len(subwayrides)
originsubway = np.sum(subwayrides * couplingsubway) / nsubway
return originpop, originwork, originsubway
def getorigin(ilat, ilong, popemp, mbta, zipscale, subwayscale):
"""
Compute features for where people live, work, and ride the subway.
"""
# locations of where people live
latbyzip = popemp['latitude'].values
longbyzip = popemp['longitude'].values
# locations of where people work
popbyzip = popemp['HD01'].values
workbyzip = popemp['EMP'].values
# replace nans with average value
finite = workbyzip * 0 == 0
mwork = workbyzip[finite].mean()
nans = workbyzip * 0 != 0
workbyzip[nans] = mwork
# locations of where people ride the subway
latbysubway = mbta['latitude']
longbysubway = mbta['longitude']
subwayrides = mbta['ridesperday'].values
# get the distance from given location to all zip codes
distancevec = densitymetric.distvec(latbyzip, longbyzip, ilat, ilong)
# compute the coupling factor to all zip codes
couplingzip = densitymetric.zipcouple(distancevec, zipscale)
# use the weighted sum as the origin score
# normalize by number of zip codes
nzip = len(workbyzip)
originpop = np.sum(popbyzip * couplingzip) / nzip
originwork = np.sum(workbyzip * couplingzip) / nzip
# get the distance from given location to all subway stops
distancevecsubway = densitymetric.distvec(latbysubway, longbysubway, ilat,
ilong)
# coupling efficiency between this station and all subway stops
couplingsubway = densitymetric.zipcouple(distancevecsubway, subwayscale)
# use the weighted sum as the origin score
# normalize by number of subway stops
nsubway = len(subwayrides)
originsubway = np.sum(subwayrides * couplingsubway) / nsubway
return originpop, originwork, originsubway
def getdestination(ilat, ilong, station, stationscale, zipscale,
stationfeatures, dataloc):
"""
Compute features associated with all possible destinations.
"""
# origin features for where people live, work and ride the subway
originpop = stationfeatures['originpop'].values
originwork = stationfeatures['originwork'].values
originsubway = stationfeatures['originsubway'].values
# location of existing stations
stationlat = station['lat'].values
stationlong = station['lng'].values
# compute the distance in miles from input station to all existing stations
distancevec = densitymetric.distvec(stationlat, stationlong, ilat, ilong)
# compute the station to station coupling factor
stationcoupling = densitymetric.stationcouple(distancevec, dataloc)
# determine coupling factor of closest station; this is used subsequently
# to compute the cannibalism factor: nrides -= nrides * maxcouple
zipcoupling = densitymetric.zipcouple(distancevec, zipscale)
maxcouple = zipcoupling.max()
# use the weighted sum as the destination score
# normalize by number of stations
norigin = len(originpop)
destpop = np.sum(originpop * stationcoupling) / norigin
destwork = np.sum(originwork * stationcoupling) / norigin
destsubway = np.sum(originsubway * stationcoupling) / norigin
return destpop, destwork, destsubway, maxcouple
def getdestination(ilat, ilong, station, stationscale, zipscale,
stationfeatures, dataloc):
"""
Compute features associated with all possible destinations.
"""
# origin features for where people live, work and ride the subway
originpop = stationfeatures['originpop'].values
originwork = stationfeatures['originwork'].values
originsubway = stationfeatures['originsubway'].values
# location of existing stations
stationlat = station['lat'].values
stationlong = station['lng'].values
# compute the distance in miles from input station to all existing stations
distancevec = densitymetric.distvec(stationlat, stationlong, ilat, ilong)
# compute the station to station coupling factor
stationcoupling = densitymetric.stationcouple(distancevec, dataloc)
# determine coupling factor of closest station; this is used subsequently
# to compute the cannibalism factor: nrides -= nrides * maxcouple
zipcoupling = densitymetric.zipcouple(distancevec, zipscale)
maxcouple = zipcoupling.max()
# use the weighted sum as the destination score
# normalize by number of stations
norigin = len(originpop)
destpop = np.sum(originpop * stationcoupling) / norigin
destwork = np.sum(originwork * stationcoupling) / norigin
destsubway = np.sum(originsubway * stationcoupling) / norigin
return destpop, destwork, destsubway, maxcouple
def getfeature(ilat, ilong, popemp, mbta, station, zipscale, stationscale,
subwayscale, stationpop, stationwork, stationsubway, latvec, longvec):
"""
Get the feature vector associated with the input latitude and longitude.
"""
stationlat = station['lat'].values
stationlong = station['lng'].values
latbyzip = popemp['latitude'].values
longbyzip = popemp['longitude'].values
popbyzip = popemp['HD01'].values
workbyzip = popemp['EMP'].values
# fix nans
finite = workbyzip * 0 == 0
mwork = workbyzip[finite].mean()
nans = workbyzip * 0 != 0
workbyzip[nans] = mwork
latbysubway = mbta['latitude']
longbysubway = mbta['longitude']
subwayrides = mbta['ridesperday'].values
distancevec = densitymetric.distvec(latbyzip, longbyzip, ilat, ilong)
couplingzip = densitymetric.coupling(distancevec, zipscale)
originpop = np.sum(popbyzip * couplingzip)
originwork = np.sum(workbyzip * couplingzip)
distancevecsubway = densitymetric.distvec(latbysubway, longbysubway,
ilat, ilong)
# coupling efficiency between this station and all subway stops
couplingsubway = densitymetric.coupling(distancevecsubway, subwayscale)
# weighted sum of subway rides
originsubway = np.sum(subwayrides * couplingsubway)
#fmt = '{0:3} {1:.5f} {2:.5f} {3:.2f} {4:.2f} {5:.5f} {6:.3f} {7:.3f}'
#print(fmt.format(ilat, ilong, originpop, originwork,
# originsubway, distancevec.min(),
# distancevec.max(), distancevec.mean()))
# test: Is there a station in stationcoupling that is ~1? Are stations
# that are known to be far from each other correctly assigned a low
# coupling efficiency? Tests indicate yes.
# compute destination scores for population, employee, and subway
destpop = []
destwork = []
destsubway = []
distancevec = densitymetric.distvec(stationlat, stationlong, ilat, ilong)
# station to station coupling
stationcoupling = densitymetric.coupling(distancevec, stationscale)
destpop = np.sum(stationpop * stationcoupling)
destwork = np.sum(stationwork * stationcoupling)
destsubway = np.sum(stationsubway * stationcoupling)
features = [originpop, originwork, destpop, destwork,
originsubway, destsubway]
maxcouple = stationcoupling.max()
return features, maxcouple
def getfeature(ilat, ilong, popemp, mbta, station, zipscale, stationscale,
subwayscale, stationpop, stationwork, stationsubway, latvec,
longvec):
"""
Get the feature vector associated with the input latitude and longitude.
"""
stationlat = station['lat'].values
stationlong = station['lng'].values
latbyzip = popemp['latitude'].values
longbyzip = popemp['longitude'].values
popbyzip = popemp['HD01'].values
workbyzip = popemp['EMP'].values
# fix nans
finite = workbyzip * 0 == 0
mwork = workbyzip[finite].mean()
nans = workbyzip * 0 != 0
workbyzip[nans] = mwork
latbysubway = mbta['latitude']
longbysubway = mbta['longitude']
subwayrides = mbta['ridesperday'].values
distancevec = densitymetric.distvec(latbyzip, longbyzip, ilat, ilong)
couplingzip = densitymetric.coupling(distancevec, zipscale)
originpop = np.sum(popbyzip * couplingzip)
originwork = np.sum(workbyzip * couplingzip)
distancevecsubway = densitymetric.distvec(latbysubway, longbysubway, ilat,
ilong)
# coupling efficiency between this station and all subway stops
couplingsubway = densitymetric.coupling(distancevecsubway, subwayscale)
# weighted sum of subway rides
originsubway = np.sum(subwayrides * couplingsubway)
#fmt = '{0:3} {1:.5f} {2:.5f} {3:.2f} {4:.2f} {5:.5f} {6:.3f} {7:.3f}'
#print(fmt.format(ilat, ilong, originpop, originwork,
# originsubway, distancevec.min(),
# distancevec.max(), distancevec.mean()))
# test: Is there a station in stationcoupling that is ~1? Are stations
# that are known to be far from each other correctly assigned a low
# coupling efficiency? Tests indicate yes.
# compute destination scores for population, employee, and subway
destpop = []
destwork = []
destsubway = []
distancevec = densitymetric.distvec(stationlat, stationlong, ilat, ilong)
# station to station coupling
stationcoupling = densitymetric.coupling(distancevec, stationscale)
destpop = np.sum(stationpop * stationcoupling)
destwork = np.sum(stationwork * stationcoupling)
destsubway = np.sum(stationsubway * stationcoupling)
features = [
originpop, originwork, destpop, destwork, originsubway, destsubway
]
maxcouple = stationcoupling.max()
return features, maxcouple
