def evaluate_station_coverage(date_val1,
date_val2,
target_station,
day_part='monrning',
max_trips=10**6):
# Scores each station by its averge wait+commute to the given target function
# INPUT:
# dateval1 - start date of trips
# dateval2 - end date of trips
# target_station_id - relative to this station all the shortest paths will be calculated
# day_part- slice the statistics as 'all', 'morning', 'noon' or 'evening'
# max_trips - limits the number of trips analyzed
trips = get_samples_by_trip(date_val1, date_val2, max_trips)
print("Got %d trip.." % (len(trips)))
G = create_train_graph(trips)
time_len = 60 * 24
hour_vec = np.array(range(24 * 60)) / 60
minuete_vec = np.array(range(24 * 60)) % 60
time_indx_real = map(lambda x: '%04d' % x, hour_vec * 100 + minuete_vec)
wait_vec, path_vec = dynamic_all_to_one(G, target_station, time_len)
station_scores = OrderedDict()
morning_indx = [7 * 60, 12 * 60]
noon_indx = [12 * 60, 17 * 60]
evening_indx = [17 * 60, 22 * 60]
max_wait_val = 120.0
for station in wait_vec:
wait_vals = wait_vec[station]
if day_part == 'all':
wait_vals = filter(lambda x: not np.isinf(x), wait_vals)
elif day_part == 'morning':
wait_vals = filter(lambda x: not np.isinf(x),
wait_vals[morning_indx[0]:morning_indx[1]])
elif day_part == 'noon':
wait_vals = filter(lambda x: not np.isinf(x),
wait_vals[noon_indx[0]:noon_indx[1]])
elif day_part == 'evening':
wait_vals = filter(lambda x: not np.isinf(x),
wait_vals[evening_indx[0]:evening_indx[1]])
wait_vals = np.array(list(wait_vals))
if len(wait_vals) == 0:
station_scores[station] = 0
else:
station_scores[station] = np.maximum(
1 - np.median(wait_vals) / max_wait_val, 0)
return station_scores
def evaluate_station_coverage(date_val1, date_val2, target_station, day_part='monrning', max_trips=10**6):
# Scores each station by its averge wait+commute to the given target function
# INPUT:
# dateval1 - start date of trips
# dateval2 - end date of trips
# target_station_id - relative to this station all the shortest paths will be calculated
# day_part- slice the statistics as 'all', 'morning', 'noon' or 'evening'
# max_trips - limits the number of trips analyzed
trips = get_samples_by_trip(date_val1, date_val2, max_trips)
print("Got %d trip.." % (len(trips)))
G = create_train_graph(trips)
time_len = 60 * 24
hour_vec = np.array(range(24 * 60)) / 60
minuete_vec = np.array(range(24 * 60)) % 60
time_indx_real = map(lambda x: '%04d' % x, hour_vec * 100 + minuete_vec)
wait_vec, path_vec = dynamic_all_to_one(G, target_station, time_len)
station_scores = OrderedDict()
morning_indx = [7 * 60, 12 * 60]
noon_indx = [12 * 60, 17 * 60]
evening_indx = [17 * 60, 22 * 60]
max_wait_val = 120.0
for station in wait_vec:
wait_vals = wait_vec[station]
if day_part == 'all':
wait_vals = filter(lambda x: not np.isinf(x), wait_vals)
elif day_part == 'morning':
wait_vals = filter(lambda x: not np.isinf(x), wait_vals[morning_indx[0]:morning_indx[1]])
elif day_part == 'noon':
wait_vals = filter(lambda x: not np.isinf(x), wait_vals[noon_indx[0]:noon_indx[1]])
elif day_part == 'evening':
wait_vals = filter(lambda x: not np.isinf(x), wait_vals[evening_indx[0]:evening_indx[1]])
wait_vals = np.array(list(wait_vals))
if len(wait_vals) == 0:
station_scores[station] = 0
else:
station_scores[station] = np.maximum(1 - np.median(wait_vals)/max_wait_val, 0)
return station_scores
hop_str = str(station_id1) + '_' + str(station_id2)
if not hop_str in edge_dict:
edge_dict[hop_str] = time_vec_template.copy()
edge_dict[hop_str][time_indx] = hop_time_expected
# Insert edge dictionary into graph structure
for hop in edge_dict:
station_id1 = int(hop.partition('_')[0])
station_id2 = int(hop.partition('_')[2])
# post process time_vec into FIFO constraint
time_vec = edge_dict[hop]
time_vec_proc = fifo_time(time_vec)
G.add_edge(station_id1, station_id2, w=time_vec_proc)
return G
### -------------------------------------------------------------------------------------------------------------------------------
if __name__ == "__main__":
max_trips = 10
date_val1 = datetime.datetime(2015, 5, 15)
date_val2 = datetime.datetime(2015, 6, 15)
trips = get_samples_by_trip(date_val1, date_val2, max_trips)
G = create_train_graph(trips)
print("Done")
### -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
if __name__ == '__main__':
# Create graph
# time_len = 2
# G = create_predefind_graph()
date_val1 = datetime.date(2015, 1, 1)
date_val2 = datetime.date(2015, 3, 12)
max_trips = 100
trips = get_samples_by_trip(date_val1, date_val2, max_trips)
# for ind in range(len(trips_full)):
# trips = {list(trips_full.keys())[ind] : trips_full[list(trips_full.keys())[ind]]}
print("Got %d trip.." % (len(trips)))
G = create_train_graph(trips)
target = 4600
time_len = 60 * 24
hour_vec = np.array(range(24 * 60)) / 60
minuete_vec = np.array(range(24 * 60)) % 60
time_indx_real = map(lambda x: '%04d' % x, hour_vec * 100 + minuete_vec)
# G = create_toy_graph(10, 25, time_len=time_len)
