def validate_point_process(
start_date=datetime.date(2011, 3, 1),
end_date=datetime.date(2012, 3, 31),
initial_cutoff=212,
num_validation=100,
num_pp_iter=100,
grid=250,
n_sample_per_grid=20,
domain=None,
pp_class=pp_models.SeppStochasticNn,
model_kwargs=None):
if not model_kwargs:
model_kwargs = {
'max_delta_t': 150,
'max_delta_d': 500,
'bg_kde_kwargs': {'number_nn': [101, 16]},
'trigger_kde_kwargs': {'number_nn': 15},
'remove_conincident_pairs': False,
'estimation_function': lambda x, y: estimation.estimator_exp_gaussian(x, y, 0.1, 50, frac_bg=None),
'seed': 42,
}
if start_date + datetime.timedelta(days=initial_cutoff + num_validation) > end_date:
warnings.warn("Requested number of validation runs is too large for the data size")
if domain is None:
domain = compute_chicago_region()
data, t0, cid = get_crimes_by_type(
crime_type='burglary',
start_date=start_date,
end_date=end_date,
domain=domain,
)
sepp = pp_class(data=data, **model_kwargs)
vb = validate.SeppValidationFixedModelIntegration(data, spatial_domain=domain, cutoff_t=initial_cutoff)
vb.set_sample_units(grid, n_sample_per_grid=n_sample_per_grid)
## TODO: check the number of iterations reported is as expected here
res = vb.run(time_step=1, n_iter=num_validation,
train_kwargs={'niter': num_pp_iter},
verbose=True)
return res, vb
def apply_sepp_stochastic_nn(data,
data_index,
domain,
grid_squares=None,
max_t=90,
max_d=500,
num_nn=(100, 15),
niter_training=50,
num_sample_points=10,
seed=43):
est_fun = lambda x, y: estimation.estimator_bowers_fixed_proportion_bg(
x, y, ct=1, cd=10, frac_bg=0.5)
trigger_kde_kwargs = {'strict': False, 'number_nn': num_nn[-1]}
bg_kde_kwargs = {'strict': True, 'number_nn': num_nn}
sepp = pp_models.SeppStochasticNn(data=data,
max_delta_t=max_t,
max_delta_d=max_d,
seed=seed,
estimation_function=est_fun,
trigger_kde_kwargs=trigger_kde_kwargs,
bg_kde_kwargs=bg_kde_kwargs)
vb = validate.SeppValidationFixedModelIntegration(data=data,
model=sepp,
data_index=data_index,
spatial_domain=domain,
cutoff_t=INITIAL_CUTOFF +
T0)
if grid_squares:
vb.roc.set_sample_units_predefined(grid_squares, num_sample_points)
else:
vb.set_sample_units(250, num_sample_points)
res = vb.run(time_step=1,
n_iter=100,
verbose=True,
train_kwargs={'niter': niter_training})
return res
sepp_objs = {}
model_objs = {}
res = {}
vb_objs = {}
data_dict = {}
cid_dict = {}
for k in kinds:
data, t0, cid = cad.get_crimes_from_dump('monsuru_cad_%s' % k)
# filter: day 210 is 27/9/2011, so use everything LESS THAN 211
### SeppValidationFixedModel with integration ROC sampling
b_sepp = True
sepp = pp_class(data=data, **model_kwargs)
vb = validate.SeppValidationFixedModelIntegration(
data=data,
model=sepp,
data_index=cid,
spatial_domain=poly,
cutoff_t=INITIAL_CUTOFF,
)
vb.set_sample_units(250, num_sample_points)
res[k] = vb.run(time_step=1,
n_iter=num_validation,
verbose=True,
train_kwargs={'niter': niter})
sepp_objs[k] = vb.model
def camden():
start_day_numbers = [277, 307, 337,
367] # number of days from t0 (1/3/2011)
poly = cad.get_camden_region()
qset = models.Division.objects.filter(type='cad_250m_grid')
qset = sorted(qset, key=lambda x: int(x.name))
grid_squares = [t.mpoly[0] for t in qset]
# define crime types
crime_types = {
'burglary': 3,
'robbery': 5,
'theft_of_vehicle': 6,
'violence': 1,
}
for (name, n) in crime_types.items():
print "Crime type: %s" % name
try:
data, t0, cid = cad.get_crimes_by_type(n)
t_upper = data[-1, 0]
sepp_objs = {}
vb_objs = {}
res = {}
for sdn in start_day_numbers:
vb = validate.SeppValidationFixedModelIntegration(
data=data,
pp_class=pp_models.SeppStochasticNn,
data_index=cid,
spatial_domain=poly,
cutoff_t=sdn,
model_kwargs=model_kwargs,
)
vb.set_sample_units(grid_squares, num_sample_points)
try:
res[sdn] = vb.run(time_step=1,
t_upper=t_upper,
verbose=True,
train_kwargs={'niter': niter})
sepp_objs[sdn] = vb.model
vb_objs[sdn] = vb
except Exception as exc:
print exc
res[sdn] = None
sepp_objs[sdn] = None
vb_objs[sdn] = None
with open(
os.path.join(ROOT_DIR, 'camden', 'model_ageing', name,
'sepp_obj.pickle'), 'w') as f:
dill.dump(sepp_objs, f)
with open(
os.path.join(ROOT_DIR, 'camden', 'model_ageing', name,
'validation_obj.pickle'), 'w') as f:
dill.dump(vb_objs, f)
with open(
os.path.join(ROOT_DIR, 'camden', 'model_ageing', name,
'validation.pickle'), 'w') as f:
dill.dump(res, f)
except Exception as exc:
with open(
os.path.join(ROOT_DIR, 'camden', 'model_ageing', name,
'errors'), 'a') as f:
f.write(repr(exc))
f.write('\n')
def run_chicago():
start_date = datetime.datetime(2011, 3, 1)
end_date = start_date + datetime.timedelta(days=277 + 480)
start_day_numbers = [277 + 30 * i for i in range(18)]
poly = chicago.compute_chicago_region()
south = models.ChicagoDivision.objects.get(name='South').mpoly
# define crime types
crime_types = {
'burglary': 'burglary',
'robbery': 'robbery',
'theft_of_vehicle': 'motor vehicle theft',
'violence': 'assault',
}
for (name, pt) in crime_types.items():
print "Crime type: %s" % name
base_dir = os.path.join(ROOT_DIR, 'chicago', 'model_ageing', name)
try:
data, t0, cid = chicago.get_crimes_by_type(crime_type=pt,
start_date=start_date,
end_date=end_date,
domain=south)
t_upper = data[-1, 0]
sepp_objs = {}
vb_objs = {}
res = {}
for sdn in start_day_numbers:
vb = validate.SeppValidationFixedModelIntegration(
data=data,
pp_class=pp_models.SeppStochasticNn,
data_index=cid,
spatial_domain=south,
cutoff_t=sdn,
model_kwargs=model_kwargs,
)
vb.set_sample_units(250, num_sample_points)
try:
res[sdn] = vb.run(time_step=1,
t_upper=t_upper,
verbose=True,
train_kwargs={'niter': niter})
sepp_objs[sdn] = vb.model
vb_objs[sdn] = vb
except Exception as exc:
print exc
res[sdn] = None
sepp_objs[sdn] = None
vb_objs[sdn] = None
if not os.path.isdir(base_dir):
os.makedirs(base_dir)
with open(os.path.join(base_dir, 'sepp_obj.pickle'), 'w') as f:
dill.dump(sepp_objs, f)
with open(os.path.join(base_dir, 'validation_obj.pickle'),
'w') as f:
dill.dump(vb_objs, f)
with open(os.path.join(base_dir, 'validation.pickle'), 'w') as f:
dill.dump(res, f)
except Exception as exc:
with open(os.path.join(base_dir, 'errors'), 'a') as f:
f.write(repr(exc))
f.write('\n')
def run_me(data, domain, out_dir, run_name, pp_class):
# data_file = os.path.join(IN_DIR, location_dir, '%s.pickle' % crime_type)
# out_dir = os.path.join(OUT_DIR, location_dir, OUT_SUBDIR)
log_file = os.path.join(out_dir, '%s.log' % run_name)
if not os.path.isdir(out_dir):
try:
os.makedirs(out_dir)
except OSError:
# wait a moment, just in case another process has just done the folder creation
time.sleep(1)
if not os.path.isdir(out_dir):
raise
# set loggers
logger = logging.getLogger(run_name)
logger.setLevel(logging.DEBUG)
fh = logging.FileHandler(log_file)
formatter = logging.Formatter(
'%(asctime)s - %(name)s - %(levelname)s - %(message)s')
fh.setFormatter(formatter)
logger.addHandler(fh)
for t in ('point_process.models', 'validation.validation'):
this_logger = logging.getLogger(t)
this_logger.setLevel(logging.DEBUG)
this_logger.handlers = [] # clear existing handlers
this_logger.addHandler(fh) # replace with the same file output
logger.info("Logger set. Script started.")
# check that num_validation iterations is feasible
if start_day_number + num_validation - 1 > data[-1, 0]:
this_num_validation = int(data[-1, 0]) - start_day_number + 1
logger.info("Can only do %d validation runs" % this_num_validation)
else:
this_num_validation = num_validation
logger.info("Instantiating validation object")
sepp = pp_class(data=data, **model_kwargs)
vb = validate.SeppValidationFixedModelIntegration(
data=data,
model=sepp,
spatial_domain=domain,
cutoff_t=start_day_number,
)
logger.info("Setting validation grid")
vb.set_sample_units(grid_size, num_sample_points)
file_stem = os.path.join(out_dir, run_name)
try:
logger.info("Starting validation run.")
res = vb.run(time_step=1,
n_iter=this_num_validation,
verbose=True,
train_kwargs={'niter': niter},
pred_kwargs={'include': pred_include})
except Exception as exc:
logger.error(repr(exc))
res = None
finally:
logger.info("Saving results (or None).")
with open(file_stem + '-validation.pickle', 'w') as f:
pickle.dump(res, f)
def chicago_south_side(min_bandwidth, crime_type):
data_file = os.path.join(IN_DIR, 'chicago_south', '%s.pickle' % crime_type)
poly_file = os.path.join(IN_DIR, 'boundaries.pickle')
out_dir = os.path.join(OUT_DIR, 'chicago_south',
'min_bandwidth_trigger_only')
log_file = os.path.join(
out_dir,
crime_type + '-' + '-'.join(['%.2f' % t
for t in min_bandwidth]) + '.log')
if not os.path.isdir(out_dir):
os.makedirs(out_dir)
# set loggers
logger = logging.getLogger('vary_min_bandwidths.chicago_south_side')
logger.setLevel(logging.DEBUG)
fh = logging.FileHandler(log_file)
formatter = logging.Formatter(
'%(asctime)s - %(name)s - %(levelname)s - %(message)s')
fh.setFormatter(formatter)
logger.addHandler(fh)
for t in ('point_process.models', 'validation.validation'):
this_logger = logging.getLogger(t)
this_logger.setLevel(logging.DEBUG)
this_logger.handlers = [] # clear existing handlers
this_logger.addHandler(fh) # replace with the same file output
logger.info("Logger set. Script started.")
logger.info("Crime type: %s. Min bandwidths %s" %
(crime_type, str(min_bandwidth)))
# load data
with open(data_file, 'r') as f:
data = pickle.load(f)
with open(poly_file, 'r') as f:
boundaries = pickle.load(f)
south_side_poly = boundaries['chicago_south']
logger.info("Loaded data.")
# check that num_validation iterations is feasible
if start_day_number + num_validation - 1 > data[-1, 0]:
this_num_validation = int(data[-1, 0]) - start_day_number + 1
logger.info("Can only do %d validation runs" % this_num_validation)
else:
this_num_validation = num_validation
model_kwargs['trigger_kde_kwargs']['min_bandwidth'] = [
min_bandwidth[0], min_bandwidth[1], min_bandwidth[1]
]
# model_kwargs['bg_kde_kwargs']['min_bandwidth'] = model_kwargs['trigger_kde_kwargs']['min_bandwidth']
logger.info("Instantiating validation object")
vb = validate.SeppValidationFixedModelIntegration(
data=data,
pp_class=pp_models.SeppStochasticNn,
spatial_domain=south_side_poly,
cutoff_t=start_day_number,
model_kwargs=model_kwargs,
)
logger.info("Setting validation grid")
vb.set_sample_units(250, num_sample_points)
try:
logger.info("Starting validation run.")
res = vb.run(time_step=1,
n_iter=this_num_validation,
verbose=True,
train_kwargs={'niter': niter})
except Exception as exc:
logger.error(repr(exc))
raise exc
else:
logger.info("Saving results.")
file_stem = os.path.join(
out_dir,
crime_type + '_' + '-'.join(['%.2f' % t for t in min_bandwidth]))
with open(file_stem + '-validation.pickle', 'w') as f:
pickle.dump(res, f)
with open(file_stem + '-vb_obj.pickle', 'w') as f:
pickle.dump(vb, f)
def camden():
# start_date is the FIRST DAY OF THE PREDICTION
start_date = datetime.datetime(2011, 12, 3)
# equivalent in number of days from t0 (1/3/2011)
start_day_number = 277
# start_day_number = 385
num_validation = 120
min_t_bds = [0, 0.5, 1, 2]
min_d_bds = [0, 20, 50, 100]
tt, dd = np.meshgrid(min_t_bds, min_d_bds)
poly = cad.get_camden_region()
qset = models.Division.objects.filter(type='cad_250m_grid')
qset = sorted(qset, key=lambda x: int(x.name))
grid_squares = [t.mpoly[0] for t in qset]
# define crime types
crime_types = {
'burglary': 3,
'robbery': 5,
'theft_of_vehicle': 6,
'violence': 1,
}
for (name, n) in crime_types.items():
print "Crime type: %s" % name
base_dir = os.path.join(ROOT_DIR, 'camden', 'min_bandwidth', name)
if not os.path.isdir(base_dir):
os.makedirs(base_dir)
try:
data, t0, cid = cad.get_crimes_by_type(n)
sepp_objs = {}
vb_objs = {}
res = {}
# jiggle grid-snapped points
# data = spatial.jiggle_on_grid_points(data, grid_squares)
# check that num_validation iterations is feasible
if start_day_number + num_validation - 1 > data[-1, 0]:
this_num_validation = int(data[-1, 0]) - start_day_number + 1
print "Can only do %d validation runs" % this_num_validation
else:
this_num_validation = num_validation
for t, d in zip(tt.flat, dd.flat):
model_kwargs['trigger_kde_kwargs']['min_bandwidth'] = [t, d, d]
model_kwargs['bg_kde_kwargs']['min_bandwidth'] = [t, d, d]
vb = validate.SeppValidationFixedModelIntegration(
data=data,
pp_class=pp_models.SeppStochasticNn,
data_index=cid,
spatial_domain=poly,
cutoff_t=start_day_number,
model_kwargs=model_kwargs,
)
vb.set_sample_units(grid_squares, num_sample_points)
try:
res[(t, d)] = vb.run(time_step=1,
n_iter=this_num_validation,
verbose=True,
train_kwargs={'niter': niter})
sepp_objs[(t, d)] = vb.model
vb_objs[(t, d)] = vb
except Exception as exc:
print exc
res[(t, d)] = None
sepp_objs[(t, d)] = None
vb_objs[(t, d)] = None
with open(os.path.join(base_dir, 'sepp_obj.pickle'), 'w') as f:
dill.dump(sepp_objs, f)
with open(os.path.join(base_dir, 'validation_obj.pickle'),
'w') as f:
dill.dump(vb_objs, f)
with open(os.path.join(base_dir, 'validation.pickle'), 'w') as f:
dill.dump(res, f)
except Exception as exc:
with open(os.path.join(base_dir, 'errors'), 'a') as f:
f.write(repr(exc))
f.write('\n')
def run_me(location_dir, location_poly, max_delta_t, max_delta_d, crime_type):
data_file = os.path.join(IN_DIR, location_dir, '%s.pickle' % crime_type)
out_dir = os.path.join(OUT_DIR, location_dir, 'max_triggers')
log_file = os.path.join(out_dir, crime_type + '_' + '%d-%d.log' % (max_delta_t, max_delta_d))
if not os.path.isdir(out_dir):
os.makedirs(out_dir)
# set loggers
logger = logging.getLogger('vary_max_triggers.%s' % location_dir)
logger.setLevel(logging.DEBUG)
fh = logging.FileHandler(log_file)
formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
fh.setFormatter(formatter)
logger.addHandler(fh)
for t in ('point_process.models', 'validation.validation'):
this_logger = logging.getLogger(t)
this_logger.setLevel(logging.DEBUG)
this_logger.handlers = [] # clear existing handlers
this_logger.addHandler(fh) # replace with the same file output
logger.info("Logger set. Script started.")
logger.info("Crime type: %s. Max delta t %d, max delta d %d" % (crime_type, max_delta_t, max_delta_d))
# load data
with open(data_file, 'r') as f:
data = pickle.load(f)
logger.info("Loaded data.")
# check that num_validation iterations is feasible
if start_day_number + num_validation - 1 > data[-1, 0]:
this_num_validation = int(data[-1, 0]) - start_day_number + 1
logger.info("Can only do %d validation runs" % this_num_validation)
else:
this_num_validation = num_validation
model_kwargs['max_delta_t'] = max_delta_t
model_kwargs['max_delta_d'] = max_delta_d
logger.info("Instantiating validation object")
vb = validate.SeppValidationFixedModelIntegration(
data=data,
pp_class=pp_models.SeppStochasticNn,
spatial_domain=location_poly,
cutoff_t=start_day_number,
model_kwargs=model_kwargs,
)
logger.info("Setting validation grid")
vb.set_sample_units(250, num_sample_points)
file_stem = os.path.join(out_dir, crime_type + '_' + '%d-%d' % (max_delta_t, max_delta_d))
try:
logger.info("Starting validation run.")
res = vb.run(time_step=1, n_iter=this_num_validation, verbose=True, train_kwargs={'niter': niter})
except Exception as exc:
logger.error(repr(exc))
res = None
finally:
logger.info("Saving results (or None).")
with open(file_stem + '-validation.pickle', 'w') as f:
pickle.dump(res, f)
with open(file_stem + '-vb_obj.pickle', 'w') as f:
pickle.dump(vb, f)