![[2602.19404] One Size Fits None: Modeling NYC Taxi Trips](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19404] One Size Fits None: Modeling NYC Taxi Trips
Summary
This paper analyzes 280 million NYC taxi trips to compare tipping behaviors between traditional taxis and app-based services, revealing distinct modeling challenges.
Why It Matters
Understanding tipping patterns in NYC's evolving taxi landscape is crucial for both drivers and ride-sharing platforms. The findings highlight the limitations of universal predictive models, emphasizing the need for tailored approaches in data analysis and service optimization.
Key Takeaways
- Traditional taxi tipping is predictable, with an R² of approximately 0.72.
- App-based tipping shows randomness, with an R² of only about 0.17.
- A single predictive model for both taxi types is ineffective due to Simpson's paradox.
- Specialized models are necessary for accurate predictions in different taxi categories.
- The study underscores the impact of technology on tipping culture in NYC.
Computer Science > Machine Learning arXiv:2602.19404 (cs) [Submitted on 10 Dec 2025] Title:One Size Fits None: Modeling NYC Taxi Trips Authors:Tomas Eglinskas View a PDF of the paper titled One Size Fits None: Modeling NYC Taxi Trips, by Tomas Eglinskas View PDF HTML (experimental) Abstract:The rise of app-based ride-sharing has fundamentally changed tipping culture in New York City. We analyzed 280 million trips from 2024 to see if we could predict tips for traditional taxis versus high-volume for-hire services. By testing methods from linear regression to deep neural networks, we found two very different outcomes. Traditional taxis are highly predictable ($R^2 \approx 0.72$) due to the in-car payment screen. In contrast, app-based tipping is random and hard to model ($R^2 \approx 0.17$). In conclusion, we show that building one universal model is a mistake and, due to Simpson's paradox, a combined model looks accurate on average but fails to predict tips for individual taxi categories requiring specialized models. Comments: Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2602.19404 [cs.LG] (or arXiv:2602.19404v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2602.19404 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Tomas Eglinskas [view email] [v1] Wed, 10 Dec 2025 23:20:40 UTC (2,678 KB) Full-text links: Access Paper: View a PDF of the paper titled One Size Fits None: Mod...
