Testing Dataset for Choice-Based Airline Fleet Assignment and Schedule Design

This repository contains an additional synthetic testing dataset (based on realistic characteristics of data from a major carrier in 2016) for the paper "Choice-Based Airline Schedule Design and Fleet Assignment: A Decomposition Approach" published on Transportation Science. If you use any of the material here, please include a reference to the paper and this webpage.

Instance Summary

Flights	Itineraries	Fare Products	Markets	Fleet Types	Aircraft	Market Share
815	4,290	47,190	819	7	186	42.36%

Schema

• flight.json

  key: flight ID

  value: {"origin": origin airport; "destination": destination airport; "depttime": departure time in hhmm format; "arrtime": arrival time in hhmm format}

• product.json

  key: product ID

  value: {"fare": price of the fare product; "cabin": cabin of the fare product, {'Y': economy, 'C': business, 'F': first}; "origin": origin airport; "destination": destination airport; "demand": unconstrained demand/attractiveness value; "market": origin-destination market; "leg": list of flight IDs which this fare product uses}

• market.json

  key: market ID, origin+destination

  value: {"total_demand": total demand/attractiveness value in this origin-destination market (including competing airline); "OA_demand": competing airlines' demand/attractiveness value in this origin-destination market}

• fleet.json

  key: fleet ID

  value: {"FCAP": seats in F cabin; "CCAP": seats in C cabin; "YCAP": seats in Y cabin; "hourly_cost": hourly operating cost; "availability": number of aircraft available for this fleet type}

• Minimum Turn Time: 35 minutes

Computation Results

• Subnetwork partition (, ), see paper for notation (cf. Table 3).

  • max_profit_ratio: 0.5%
  • number of flights in : 143
    • of size: 1: 74, 2: 16, 3: 11, 4: 1
  • number of flights in : 672
    • of size: min: 4, max: 641, : 6
  • fare optimization CPU time: 31 seconds

• 5 hour CPU time, 2.3 GHz Quad-Core Intel Core i7, Gurobi 9.02, see paper for notation (cf. Table 4).

Model	Fare Split	Obj. Val.	LP Relax.	CPU Time(s)	Solver Gap	B&B Node	Profit	Num of Flights	Annual Profit Improvement (Million)
ISD-FAM	--	4,328,945	4,584,530	18,000	0.04%	30,722	6,159,056	743	0
ISD-FAM-SR-ITIN	--	3,891,445	4,140,062	18,000	4.38%	4,903	6,047,442	711	(40.74)
CSD-FAM	--	6,208,314	6,748,932	18,000	5.08%	5,992	6,208,314	662	17.98
CSD-FAM-S	--	5,955,392	6,630,624	18,000	8.05%	1,148	5,955,392	616	(74.34)
S-CSD-FAM (, )	opt	6,328,372	6,721,355	18,000	3.50%	6,936	6,255,524	679	35.21