def getDrillDown(self, df=None, orgs=None, dests=None, flights=None,
cabins=None, bcs=None, date_ranges=None):
"""
Filters a dataframe for rows that match the given features.
For instance, orgs=['DXB', 'DMM'], bcs=['B'] will return the dataframe
that contains rows with flights that departed from DXB or DMM AND that
were made in the 'B' booking class
args:
df: a pd.DataFrame object that the user wants to filter
orgs: list of strings describing origin airports
dests: list of strings describing destination airports
flights: list of ints describing flight numbers
cabins: list of strings describing cabins ('Y' or 'J')
bcs: list of strings describing booking classes
data_ranges: list of tuples of strings describing desired dates
returns:
pd.DataFrame instance
"""
if not isinstance(df, pd.DataFrame):
df = self.entities.copy()
false_mask = pd.Series(False, list(df.index))
m = pd.Series(True, list(df.index))
m = m & self._mask(false_mask, orgs, df.ORG)
m = m & self._mask(false_mask, dests, df.DES)
m = m & self._mask(false_mask, flights, df.FLT)
m = m & self._mask(false_mask, bcs, df.BC)
m = m & self._mask(false_mask, date_ranges, df.DATE)
if cabins is not None:
cs = []
for cabin in cabins:
classes = Utils.mapCabinToBookingClasses(cabin)
cs.extend([bc for bc, rank in classes])
m = m & self._mask(false_mask, cs, df.BC)
return df[m]
def bookingClassTicketFrequencies(f, data, cabin):
print "Grouping into unique flight/booking class combinations"
flight_data = f.getUniqueFlightsAndBookings(data)
bcs = Utils.mapCabinToBookingClasses(cabin)
bcs = {bc: 0 for (bc, r) in bcs}
print "Iterating through all booking classes"
for flight, flight_df in flight_data:
bc = flight[-1]
keyday = -1 * flight_df["KEYDAY"]
bkd = flight_df["BKD"]
keyday, bkd = Utils.sortByIndex(keyday, bkd)
bcs[bc] += bkd[-1]
total_bkd = 0.0
for bc, num_bkd in bcs.items():
total_bkd += num_bkd
for bc in bcs:
bcs[bc] /= total_bkd
ks, vs = zip(*bcs.items())
ks, vs = zip(*sorted(zip(ks, vs), key=lambda tup: Utils.compareBCs(tup[0])))
indices = np.arange(len(ks))
width = 0.75
fig, ax = plt.subplots()
rects = ax.bar(indices, vs, width)
ax.set_ylabel("Percent of Total Booked")
ax.set_title("Booking Class Ticketing Distribution - Economy Cabin")
ax.set_xticks(indices + width / 2.0)
ax.set_xticklabels(ks)
plt.grid()
plt.show()
def bc_bars_base(y_cumsum):
"""
args:
y_cumsum: a sumcum deltabkd vector (either predict or actual)
"""
totalbkd_vector = y_cumsum[:, -1]
assert len(ids_test) == len(totalbkd_vector)
bcs = {bc: 0 for (bc, r) in Utils.mapCabinToBookingClasses("Y")}
for totalbkd, ids in zip(totalbkd_vector, ids_test):
bc = ids[-1]
bcs[bc] += totalbkd
denom = sum(totalbkd_vector)
for bc in bcs:
bcs[bc] /= denom
ks, vs = zip(*bcs.items())
ks, vs = zip(*sorted(zip(ks, vs), key=lambda tup: Utils.compareBCs(tup[0])))
indices = np.arange(len(ks))
return ks, vs, indices
