def authCurves(self, network, orgs=None, dests=None, flights=None,
cabins=None, bcs=None, date_ranges=None):
""" Plots AUTH curves for some subset of the data.
AUTH is stated at the level of a cabin-booking class. AUTH changes
with time starting from the opening of ticket sales and ending close
to departure. Note that you only have to look at two booking
classes (BC) for purpose of overbooking: Y class for Y cabin and J
class for J cabin. This is because those classes always have the
maximum AUTH among all classes in a cabin at a given point of time
(they are at the top in hierarchy).
"""
df = network.f.getDrillDown(orgs=orgs, dests=dests, flights=flights,
cabins=cabins, bcs=bcs, date_ranges=date_ranges)
fltbk = network.f.getUniqueFlightsAndBookings(df)
plt.figure()
for g, d in fltbk:
AUTH = np.array(d.sort(columns='KEYDAY', ascending=False)['AUTH'])
KEYDAY = np.array(-d.sort(columns='KEYDAY', ascending=False)['KEYDAY'])
plt.plot(KEYDAY, AUTH)
title = Utils.createTitleForFeatures(orgs,dests,flights,cabins,bcs,date_ranges)
plt.title(title)
plt.xlabel('-KEYDAY')
plt.ylabel('AUTH')
plt.show()
def bookingCurves(self, network, orgs=None, dests=None, flights=None,
cabins=None, bcs=None, date_ranges=None):
""" Plots booking curves for some subset of the data.
A booking curve tracks the number of seats booked over time, starting
from the opening of ticket sales and ending close to departure
"""
df = network.f.getDrillDown(orgs=orgs, dests=dests, flights=flights,
cabins=cabins, bcs=bcs, date_ranges=date_ranges)
fltbk = network.f.getUniqueFlightsAndBookings(df)
plt.figure()
for g, d in fltbk:
BKD = list(d.sort(columns='KEYDAY', ascending=False)['BKD'])
KEYDAY = list(-d.sort(columns='KEYDAY', ascending=False)['KEYDAY'])
ID = d['DATE'].first
BC = d['BC'].first
plt.plot(KEYDAY, BKD)
title = Utils.createTitleForFeatures(orgs,dests,flights,cabins,bcs,date_ranges)
plt.title(title)
plt.xlabel('-KEYDAY')
plt.ylabel('BKD')
plt.show()
def overbookingCurves(self, network, orgs=None, dests=None, flights=None,
cabins=None, bcs=None, date_ranges=None, normalized=True):
""" Plots overbooking curves for some subset of the data.
Overbooking is defined where AUTH > CAP. We plot overbooking as a
ratio between AUTH and CAP. Overbooking varies with time.
"""
df = network.f.getDrillDown(orgs=orgs, dests=dests, flights=flights,
cabins=cabins, bcs=bcs, date_ranges=date_ranges)
fltbk = network.f.getUniqueFlightsAndBookings(df)
plt.figure()
if normalized:
for g, d in fltbk:
# normalized AUTH == OVERBOOKED
AUTH = np.array(d.sort(columns='KEYDAY', ascending=False)['AUTH'])
# ignore time series that are not overbooked
if not Utils.isOverbooked(AUTH):
continue
KEYDAY = np.array(-d.sort(columns='KEYDAY', ascending=False)['KEYDAY'])
plt.plot(KEYDAY, AUTH)
else:
for g, d in fltbk:
AUTH = np.array(d.sort(columns='KEYDAY', ascending=False)['AUTH'])
CAP = float(d.iloc[0]['CAP'])
OVRBKD = AUTH/CAP
# ignore time series that are not overbooked
if not Utils.isOverbooked(OVRBKD):
continue
KEYDAY = np.array(-d.sort(columns='KEYDAY', ascending=False)['KEYDAY'])
plt.plot(KEYDAY, OVRBKD)
title = Utils.createTitleForFeatures(orgs,dests,flights,cabins,bcs,date_ranges)
plt.title(title)
plt.xlabel('-KEYDAY')
plt.ylabel('Percentage Overbooked: AUTH / CAP')
plt.show()
def stackedBookingCurve(self, network, orgs=None, dests=None,
flights=None, cabins=None, bcs=None,
date_ranges=None):
"""
Generate a summative booking curve for a given flight. In order for this
function to work properly the arguments must specify one specific flight
(or a subset of the booking classes on a specific flight). Additionally,
the network must have been create using a normalized data set.
"""
first_flights = network.f.getDrillDown(orgs=orgs, dests=dests,
flights=flights, cabins=cabins,
bcs=bcs, date_ranges=date_ranges)
groupedByBookings = network.f.getUniqueFlightsAndBookings(first_flights)
xvals = np.linspace(-1, 0, 101) # Increments of .01 from -1 -> 0
interps = None
labels = [g[4] for g, d in groupedByBookings]
for g, d in groupedByBookings:
keydays = -d['KEYDAY']
booked = d['BKD']
yvals = network.interp(xvals, keydays, booked)
if interps == None:
interps = yvals
else:
interps = np.vstack((interps, yvals))
# interps is my matrix
m, n = interps.shape
interps_sum = np.zeros((m,n))
for i in range(m-1):
for j in range(i+1, m):
interps_sum[j] += interps[i]
for i in range(m):
plt.plot(xvals, interps_sum[i])
plt.legend(labels, loc=6, prop={'size': 14})
plt.title('Summative Booking Curve\n' + Utils.createTitleForFeatures(orgs, dests, flights, cabins, bcs, date_ranges))
plt.xlabel('Normalized Keyday')
plt.ylabel('Normalized Booked')
plt.show()
def overbookingVsCabinLoadFactor(self, network, orgs=None, dests=None, flights=None,
cabins=None, bcs=None, date_ranges=None,
normalized=True, subplots=True):
""" Plots how overbooking varies with Cabin load factor. Final Cabin Load Factor
for a particular flight booking class is binned into three separate categories:
Overbooked: CLF > 1
Underbooked: CLF < .8
Optimumly booked: .8 < CLF < 1
"""
df = network.f.getDrillDown(orgs=orgs, dests=dests, flights=flights,
cabins=cabins, bcs=bcs, date_ranges=date_ranges)
fltbk = network.f.getUniqueFlightsAndBookings(df)
# TODO: allow for countFinalCabinLoadFactor to use normalized data
CLF_dict = network.countFinalCabinLoadFactor()
fig = plt.figure()
# preparing to capture the legend handles
legend_over = None
legend_under = None
legend_optimum = None
n_over = 0
n_under = 0
n_optimum = 0
if normalized:
for g, d in fltbk:
# normalized AUTH == OVERBOOKED
AUTH = np.array(d.sort(columns='KEYDAY', ascending=False)['AUTH'])
# ignore time series that are not overbooked
if not Utils.isOverbooked(AUTH):
continue
KEYDAY = np.array(-d.sort(columns='KEYDAY', ascending=False)['KEYDAY'])
DATE = d.iloc[0]['DATE']
FLT = d.iloc[0]['FLT']
ORG = d.iloc[0]['ORG']
DES = d.iloc[0]['DES']
#TODO: See CLF_dict (above)
CABIN_LOAD_FACTOR = CLF_dict[(DATE, FLT, ORG, DES)]
if CABIN_LOAD_FACTOR > 1:
plt.plot(KEYDAY, AUTH, 'r')
elif CABIN_LOAD_FACTOR < .95:
plt.plot(KEYDAY, AUTH, 'y')
else:
plt.plot(KEYDAY, AUTH, 'g')
else:
for g, d in fltbk:
AUTH = np.array(d.sort(columns='KEYDAY', ascending=False)['AUTH'])
CAP = float(d.iloc[0]['CAP'])
OVRBKD = AUTH/CAP
# ignore time series that are not overbooked
if not Utils.isOverbooked(OVRBKD):
continue
KEYDAY = np.array(-d.sort(columns='KEYDAY', ascending=False)['KEYDAY'])
DATE = d.iloc[0]['DATE']
FLT = d.iloc[0]['FLT']
ORG = d.iloc[0]['ORG']
DES = d.iloc[0]['DES']
CABIN_LOAD_FACTOR = CLF_dict[(DATE, FLT, ORG, DES)]
if CABIN_LOAD_FACTOR > 1:
plt.subplot(311) if subplots else None
if not legend_over:
legend_over, = plt.plot(KEYDAY, OVRBKD , 'r')
else:
plt.plot(KEYDAY, OVRBKD , 'r')
n_over += 1
elif CABIN_LOAD_FACTOR < .95:
plt.subplot(313) if subplots else None
if not legend_under:
legend_under, = plt.plot(KEYDAY, OVRBKD, 'y')
else:
plt.plot(KEYDAY, OVRBKD, 'y')
n_under += 1
else:
plt.subplot(312) if subplots else None
if not legend_optimum:
legend_optimum, = plt.plot(KEYDAY, OVRBKD, 'g')
else:
plt.plot(KEYDAY, OVRBKD, 'g')
n_optimum += 1
title = Utils.createTitleForFeatures(orgs,dests,flights,cabins,bcs,date_ranges)
plt.subplot(311) if subplots else None
plt.suptitle(title)
plt.xlabel('-KEYDAY')
plt.ylabel('Percentage Overbooked: AUTH / CAP')
plt.subplot(311)
plt.ylim([.25,2])
plt.subplot(312)
plt.ylim([.25,2])
plt.subplot(313)
plt.ylim([.25,2])
plt.legend([legend_over, legend_optimum, legend_under],
["Cabin Load Factor > 1, n={}".format(n_over),
"Optimum Cabin Load Factor,n={}".format(n_optimum),
"Cabin Load Factor < .95, n={}".format(n_under)
])
plt.show()
