def solve_iteratively(Config, region_octtree, regions, raster, raster_affine):
##
# if num zones is too large, we need a higher threshold
# keep a record of the thresholds that result in the nearest low, and nearest high
# for the next step, take the halfway number between the two
desired_num_zones = Config.getint("Parameters", "desired_num_zones")
best_low = Config.getint("Parameters", "lower_population_threshold")
best_high = Config.getint("Parameters", "upper_population_threshold")
tolerance = Config.getfloat("Parameters", "tolerance")
times = []
if best_low is None or best_low == 0: best_low = 1
if best_high is None or best_high == 0: best_high = np.sum(raster)
step = 1
solved = False
num_zones = 0
pop_threshold = (best_high - best_low) / 2
start_time = time.time()
times += [[0, 0]]
while not solved: # difference greater than 10%
print 'step %d with threshold level %d...' % (step, pop_threshold)
prev_num_zones = num_zones
region_octtree = build_out_nodes(Config, region_octtree, regions, raster, raster_affine, pop_threshold)
num_zones = region_octtree.count_populated()
print "\tnumber of cells:", num_zones
print ''
solved = prev_num_zones == num_zones or abs(num_zones - desired_num_zones)/float(desired_num_zones) < tolerance
if not solved:
if num_zones > desired_num_zones:
best_low = max (best_low, pop_threshold)
else:
best_high = min (best_high, pop_threshold)
pop_threshold = (best_low + best_high) / 2
step += 1
times += [[time.time()-start_time, num_zones]]
print "Solution found!"
print "\t%6d zones" % (num_zones)
print "\t%6d threshold" % (pop_threshold)
with open('output/time.csv', 'wb') as time_csv:
fieldnames = ['time', 'zones']
timewriter = csv.writer(time_csv, delimiter=',',
quotechar='"', quoting=csv.QUOTE_MINIMAL)
timewriter.writerow(fieldnames)
for x in times:
timewriter.writerow(x)
return region_octtree
def model_zones_vs_threshold(Config, region_octtree, regions, raster, raster_affine):
print 'running trend analysis...'
results = []
for pop_threshold in xrange(2000, 20000, 2000):
octtree = build_out_nodes(Config, region_octtree, regions, raster, raster_affine, pop_threshold, split=False)
num_zones = octtree.count_populated()
results.append((pop_threshold, num_zones))
print results[-1]
from matplotlib import pyplot
pyplot.plot(*zip(*results))
pyplot.xlabel('Data Threshold used (population + employment)')
pyplot.ylabel('Number of zones created')
pyplot.title('Trends of zone size')
pyplot.show()
def solve_iteratively(Config, region_octtree, regions, raster, raster_affine):
##
# if num zones is too large, we need a higher threshold
# keep a record of the thresholds that result in the nearest low, and nearest high
# for the next step, take the halfway number between the two
desired_num_zones = Config.getint("Parameters", "desired_num_zones")
best_low = Config.getint("Parameters", "lower_population_threshold")
best_high = Config.getint("Parameters", "upper_population_threshold")
tolerance = Config.getfloat("Parameters", "tolerance")
step = 1
solved = False
num_zones = 0
pop_threshold = (best_high - best_low) / 2
while not solved: # difference greater than 10%
print 'step %d with threshold level %d...' % (step, pop_threshold)
region_octtree = build_out_nodes(Config, region_octtree, regions, raster, raster_affine, pop_threshold)
num_zones = region_octtree.count_populated()
print "\tnumber of cells:", num_zones
print ''
solved = abs(num_zones - desired_num_zones)/float(desired_num_zones) < tolerance
if not solved:
if num_zones > desired_num_zones:
best_low = max (best_low, pop_threshold)
else:
best_high = min (best_high, pop_threshold)
pop_threshold = (best_low + best_high) / 2
step += 1
print "Solution found!"
print "\t%6d zones" % (num_zones)
print "\t%6d threshold" % (pop_threshold)
return region_octtree
