A Two-Stage Hybrid Framework for Inferring Air Trips from Mobile Location-based Service (LBS) Data under Incomplete Airport Sightings
Authors: Hsuan-Chih Wang, Xianfeng Terry Yang
Organizations: Department of Civil and Environmental Engineering, University of Maryland College Park, MD, 20742 · Maryland Transportation Institute Department of Civil and Environmental Engineering, University of Maryland College Park, MD, 20742
Mobile device location-based service (LBS) data provide large-scale spatiotemporal observations that can help airport planners understand air travelers' mobility patterns. However, airport sightings are often missing before or after flights, making strictly rule-based detection incomplete. This study develops a two-stage framework to infer U.S. domestic air trips from large-scale LBS data under incomplete airport sightings. Stage 1 uses heuristic rules to identify high-confidence air trips whose endpoints are observed at two distinct airports. Stage 2 addresses latent air trips with a probabilistic model that infers air routes using spatial proximity, temporal compatibility, and historical flight schedule information. Using January 2022 nationwide LBS data with nearly 10 million users and 131 million sightings, results are validated against T-100 Domestic Segment route benchmarks and a Stage 1 pseudo-ground-truth reference set. The framework increases air route coverage from 10.65% under Stage 1 to 50.60% and recovers nearly all top-1,000 benchmark routes. It improves Spearman rank correlation with T-100 route shares from 0.4490 to 0.7630 and reduces Jensen-Shannon divergence from 0.2808 to 0.2263. Stage 2 also recovers 92.2% of Stage 1 reference trips. This study formalizes three endpoint observation regimes for LBS-based air trip inference (airport-observed, semi-observed, and latent) and helps airport agencies identify domestic air travel patterns missed when airport sightings are incomplete.
Figures & tables
Figure 1: Three endpoint-sighting regimes in LBS-based air-trip inference and their corresponding modeling strategies.
Figure 2: Overview of the proposed two-stage hybrid inference framework
Variables
u
An anonymized mobile-device user.
eu
Candidate air-trip segment for user u .
xo , xd
Observed pre-transition and post-transition sighting locations.
to , td
Timestamps associated with xo and xd , respectively.
O
Observed spatial state, O=(xo,xd) .
ϕ , λ
Latitude and longitude of a location sighting.
Table 1: List of Key Notations
ID
Timestamp
Latitude
Longitude
Place name
Sfacd-243rg
2022-01-07 00:10:09
26.069201
-80.14125
convenient store
Sfacd-243rg
2022-01-07 00:45:10
26.069323
-80.14297
Fort Lauderdale-Hollywood International Airport
Sfacd-243rg
2022-01-07 00:49:35
26.069323
-80.14297
Fort Lauderdale-Hollywood International Airport
Sfacd-243rg
2022-01-07 07:31:07
29.986477
-95.34981
Houston - George Bush Intercontinental Airport (IAH)
Sfacd-243rg
2022-01-07 07:33:14
29.986478
-95.34982
Houston - George Bush Intercontinental Airport (IAH)
Sfacd-243rg
2022-01-07 11:21:20
29.986563
-95.34812
chain coffee shop
Table 2: Sample of Mobile Device Location Data
Parameter
Symbol
Value
Maximum search radius
ξ
50 miles
Origin/destination supply weight
α
1.0
Distance friction weight
β
1.0
Temporal tolerance
σ
120 min
Cruise speed
Vc
500 mph
Time buffer
Tb
120 min
Table 3: Parameter settings used in the probabilistic inference model
Model
Identified unique air routes
Actual air routes from T-100 data
Coverage
Stage 1 only
995
9,339
10.65%
Proposed framework
4,726
9,339
50.60%
Table 4: Route-level detection performance against T-100 Domestic Segment data
Air route group
Stage 1
Proposed framework
Top 20 routes
19 / 20 (95.00%)
20 / 20 (100.00%)
Top 50 routes
46 / 50 (92.00%)
48 / 50 (96.00%)
Top 100 routes
88 / 100 (88.00%)
98 / 100 (98.00%)
Top 500 routes
357 / 500 (71.40%)
486 / 500 (97.20%)
Top 1,000 routes
556 / 1,000 (55.60%)
973 / 1,000 (97.40%)
All recovered air routes
995 / 9,339 (10.65%)
4,726 / 9,339 (50.60%)
Table 5: Top- N route coverage against T-100 Domestic Segment data
Metric
Stage 1
Proposed framework
Spearman correlation
0.4490
0.7630
Share MAE
0.0123%
0.0117%
JS divergence
0.2808
0.2263
Table 6: Route-share distribution metrics against T-100 Domestic Segment data
Figure 3: Recovery of the historical top-1,000 domestic air-route network
Metric
Value
Stage 1 unique air-trip count
2,073
Stage 2 unique air-trip count
305,563
Overlapped unique air-trip count
1,911
Stage 1 trips missed by Stage 2
162
Stage 2-only trips
303,490
Coverage of Stage 1 trips by Stage 2
92.2%
Table 7: Internal consistency of Stage 2 against Stage 1 pseudo-ground-truth reference set
Incomplete mobility trajectories can obscure the observations needed to identify important travel events and their endpoints. This study evaluates whether compact, locally deployed large language models (LLMs) can recover missing mobility information from incomplete trajectories, using high-confidence air trips as a controlled test case, and whether more explicit prompts can compensate for the loss of endpoint evidence. Using nationwide location-based services (LBS) data from the United States in January 2022 and airport reference data, thirty high-confidence air trips were converted into three incomplete-trajectory conditions by removing observations near the origin airport, the destination airport, or both. Four prompt variants and two LLMs generated 720 requests, each using a two-stage procedure to identify the flight transition segment and then select the origin and destination airports. Exact transition accuracy is 9.2%, while origin and destination accuracy are 31.4% and 10.7%, respectively. Only 1.9% of requests recover both the rule-identified transition and the reference airport pair. The most explicit prompt produced the highest observed exact transition and exact airport-pair recovery rates. The study introduces a controlled endpoint-removal framework for evaluating recovery of discrete mobility events from incomplete trajectories, allowing failures to be examined by inference stage and helping analysts decide whether recovered information should be retained, independently verified, or left unresolved.
Hsuan-Chih Wang, Xianfeng Terry Yang
Department of Civil and Environmental Engineering, University of Maryland, College Park, MD, 20742 · Maryland Transportation Institute
Large-scale GPS trajectory data offer rich observations of human mobility, yet assigning trip purposes to detected stops remains challenging due to the absence of individual-level ground truth, spatial uncertainty from GPS noise and incomplete points of interest (POIs) coverage, and fundamental behavioral differences across trip purposes. We propose a weakly supervised framework integrating neighborhood-level POI semantic zones with distance-weighted spatial likelihoods, differentiated inference strategies for mandatory and non-mandatory activities, and a multi-phase Pareto optimization that jointly minimizes distributional divergence from household travel survey statistics and maximizes inference reliability without requiring annotated labels. Evaluated on over 81 million staypoints in Los Angeles, the framework reduces activity type frequency Jensen-Shannon distance (JSD) by 23%, start time JSD by 48%, and duration JSD by 12% respectively relative to a comparable baseline. The proposed approach provides a scalable and uncertainty-aware path from raw GPS trajectories to semantically annotated mobility data for travel demand modeling and transportation policy analysis.
Bo Yang, Haoxuan Ma, Yifan Liu +4
1UCLA Mobility Lab, Department of Civil and Environmental Engineering, University of California, Los Angeles, Los Angeles, USA. · 2Novateur Research Solutions, Ashburn, VA, USA.
In this paper, we propose the Rule-based Visiting Circulation (RVC) model in tackling the challenge in the IEEE Big Data Cup 2022: Trip Destination Prediction. Given trips containing travel information, personal attributes, origin zones, and their features in the training metropolitan areas, the task is to predict the destination of every trip in a targeted metropolitan area whose destinations are not given at all at the training stage. We highlight the challenges in this destination prediction task -- having no knowledge of the destinations in the targeted metropolitan area. We provide insights from the datasets, in which revisiting behaviors and the relationships between origins and destinations play a crucial role in individuals' trips. Hence, we design a simple but comprehensive method, rule-based visiting circulation, which directly utilizes the origin information and individuals' trip behaviors to determine the destinations in the targeted metropolitan area, i.e., requiring no learning from the four training areas. Experimental results on both offline evaluation and leaderboard submission consistently exhibit the proposed RVC can significantly outperform supervised learning methods and other heuristics. The RVC method eventually brings us to second place in the competition leaderboard.