Geospatial and Temporal Trends in Urban Transportation: A Study of NYC Taxis and Pathao Food Deliveries

Bidyarthi Paul; Dipta Biswas; Fariha Tasnim Chowdhury; Meherin Sultana

REVIEW 4 major objections 6 minor 2 cited by

Geospatial and Temporal Trends in Urban Transportation: A Study of NYC Taxis and Pathao Food Deliveries

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A weekly-season SARIMAX model fit to one month of NYC taxi pickups is claimed to forecast daily demand well enough to guide fleet scheduling.

desk verdict A modest EDA case study whose forecast claim outruns its data; the descriptive parts are fine, the SARIMAX evidence is not. read the letter →

arxiv 2505.03816 v1 pith:R7LBHOE6 submitted 2025-05-02 cs.SI cs.AI

Bidyarthi Paul , Fariha Tasnim Chowdhury , Dipta Biswas , Meherin Sultana This is my paper · ORCID

classification cs.SIcs.AI

keywords urbantransportationtaxidemandforecastingSARIMAXNYCdatafooddeliveryweeklyseasonalitygeospatialclusteringK-means

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that demand in two urban services—yellow taxis in New York City and food deliveries in Dhaka—is regular enough to summarize, map, and forecast. Its central quantitative claim is that a SARIMAX model with a weekly seasonal term, fit to daily NYC taxi pickup counts from January 2013, predicts daily demand well enough to guide fleet placement and driver scheduling, with an in-sample RMSE of 734.944 and an out-of-sample RMSE of 204.1525. The paper also claims that Manhattan leads pickup demand, Queens has the longest average trips, and Pathao food orders peak on Fridays and Saturdays. A reader should care because the same pipeline from exploratory analysis to seasonal forecasting could transfer to other cities and delivery fleets.

What carries the argument

The central object is the SARIMAX model—seasonal autoregressive integrated moving average with exogenous variables—configured with order (1,0,1) and seasonal order (1,0,1,7). The seasonal period of 7 is the mechanism that encodes the weekly cycle in daily pickup counts, and the paper's case rests on that term's ability to reduce forecast error to the reported RMSE values. Supporting machinery includes K-means clustering with 15 clusters on pickup and dropoff coordinates, which identifies geographic high- and low-demand zones, and feature engineering that extracts hour, day, day of week, and trip duration from datetime fields.

What would settle it

Fit the same SARIMAX(1,0,1)(1,0,1,7) model to several months of NYC taxi data with a holdout of at least four weeks, and compare its out-of-sample RMSE with a plain ARIMA(1,0,1) that has no seasonal term; if the seasonal model is not better, or if significant autocorrelation remains at lag 7, the weekly-seasonality claim would not survive.

Watch

Extended reading notes

Core claim

Stated on the paper's own terms, the central discovery is that daily New York City taxi demand in January 2013 carries a weekly seasonal structure that a SARIMAX(1,0,1)(1,0,1,7) model captures: the fitted model yields an in-sample RMSE of 734.944 and an out-of-sample RMSE of 204.1525 on a short test window. The authors take these numbers as evidence that the model 'effectively captured the weekly seasonality' and that it can help operators forecast demand patterns for optimized fleet management and efficient driver scheduling. The companion descriptive findings are that Manhattan dominates pickup volume, Staten Island has the least activity, Queens shows the longest average trip distance and duration, trips lengthen later in the day and on weekends, and Pathao food delivery demand peaks on Friday and Saturday.

Load-bearing premise

The central claim depends on the assumption that one month of daily trip counts—January 2013—contains enough repeated weekly cycles to estimate and test a weekly seasonal model, with a test period of only about six days.

Editorial extensions

If this is right

Taxi operators who adopt the model can shift drivers toward late-day and weekend hours where pickups and trip lengths rise, and concentrate vehicles in Manhattan while reducing idle capacity in Staten Island.
Retraining the same SARIMAX configuration on longer NYC taxi histories should produce rolling daily demand forecasts suitable for dispatch planning, if the weekly pattern holds across months.
Food delivery operators in Dhaka can act on the weekday/weekend demand split by scheduling more riders and preparing popular items on Fridays and Saturdays, even though the Pathao dataset lacks timestamps and coordinates for a full time-series model.
Positioning idle taxis near the high-demand clusters identified by K-means should cut wait times, provided those clusters remain stable over the weeks the model covers.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A testable extension the paper does not report is a head-to-head comparison with a non-seasonal ARIMA on the same holdout, which would isolate the seasonal term's marginal contribution.
Aggregating the Pathao records by day of week and hour would test whether the Friday-Saturday ordering peak persists after accounting for how many restaurants are open.
Repeating the fit on other months of 2013 would show whether the weekly seasonal order and the Manhattan/Queens/Staten Island rankings are stable, which the one-month design does not establish.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Punchline: this is a workmanlike EDA on two public datasets, but the forecasting claim is not supported by the evidence. The descriptive patterns are probably correct; the SARIMAX conclusion is not.

What the paper does well: it gives a clear, honest account of the Pathao Food dataset, which has been analyzed less often, and it openly admits in Section 7 that the one-month NYC Taxi window limits the scope. The bar charts and cluster maps appear consistent with the underlying data. For a reader who wants a straightforward template for taxi-demand EDA, this is not a bad starting point.

Soft spots: the load-bearing claim in Section 5.3 is that a (1,0,1)(1,0,1,7) SARIMAX 'effectively captured weekly seasonality' with an in-sample RMSE of 734.94. The dataset is January 2013 only, roughly 31 daily counts. The 80/20 split leaves six test points, less than one seasonal cycle. Out-of-sample RMSE (204.15) is reported but unused as evidence, and there is no baseline, no forecast horizon, no uncertainty interval. You cannot validate a period-7 seasonal process on fewer than seven test observations. The Pathao data also lack geospatial and datetime attributes, so the paper's own frame of a 'geospatial and temporal' comparison between NYC and Dhaka cannot hold; the Pathao analysis is EDA only. The K-means step uses an unexplained K=15. No code or data are provided, so nothing is reproducible.

Verdict: this is a student-scale case study, not a contribution to the literature. The NYC taxi findings replicate known patterns. It would be a reasonable teaching illustration but not a publishable research claim as written. I would not send it to a serious referee in this form. If the authors tone down the forecasting language, add baselines and error bars, publish code, and reframe the Pathao part as descriptive EDA, it might become a workshop-level paper. For now, desk reject seems right.

Referee Report

4 major / 6 minor

Summary. The paper analyzes two urban transportation datasets: NYC Taxi Trip data from January 2013 and Pathao Food Delivery data from Dhaka. It performs exploratory data analysis, borough-level geospatial summaries, K-means clustering of pickup/dropoff coordinates, and a SARIMAX time-series forecast for taxi demand. The descriptive findings are that Manhattan has the highest taxi pickup volume, Staten Island the lowest, taxi trips tend to be longer later in the day and on weekends, and Pathao orders peak on Fridays and Saturdays. The paper also claims that the SARIMAX model effectively captured weekly seasonality with an RMSE of 734.944 and that this can support fleet management and driver scheduling.

Significance. If the forecasting claim were supported, the work would be practically relevant for taxi and food-delivery operators. The descriptive parts are straightforward and generally consistent with the displayed figures, and the authors are honest in Section 7 that the single-month NYC sample limits generalizability. The strengths are the use of real-world data, the clear exploratory summaries, and the explicit acknowledgment of scope limitations. However, the time-series evidence is underpowered: a period-7 seasonal model is fitted to roughly 31 daily observations with a test window shorter than one seasonal cycle, and the headline RMSE is the in-sample value. With additional baselines, proper out-of-sample evaluation, and more cautious claims, this could become a useful applied case study, but the central forecasting claim is not currently established.

major comments (4)

[§3.1.1, §3.3, §5.3] The central forecasting claim is not supported by the reported experimental setup. The NYC data are restricted to January 2013, about 31 daily observations, and the 80:20 train-test split leaves roughly 6 test points, fewer than one full weekly cycle. A SARIMAX model with seasonal order (1,0,1,7) cannot be validated on a test window shorter than the seasonal period, and estimating seasonal AR and MA terms from roughly 25 training points is statistically fragile. The authors should either use a longer series, validate with multiple seasonal cycles, or explicitly present the single-month result as a methodological illustration rather than as evidence that weekly seasonality has been captured.
[§5.3] The RMSE reporting does not justify the conclusion drawn. The quoted RMSE of 734.944 is the in-sample value; the out-of-sample value of 204.1525 is reported but not interpreted, and no baseline model, forecast horizon, or confidence interval is provided. In-sample RMSE cannot serve as evidence of forecast skill, and the out-of-sample value needs context such as a seasonal-naive baseline or a non-seasonal ARIMA comparison. Section 5.3 should be rewritten so that the conclusion is drawn only from out-of-sample performance with suitable baselines.
[§4.4, §5.4] The clustering analysis does not support the claim that K-means identifies high- and low-demand regions. The number of clusters K=15 is chosen without a selection criterion or stability check, and the clustering input is raw latitude and longitude coordinates, not demand counts. The resulting figure likely shows geographic groupings rather than demand hotspots. To support the claim, the authors should combine cluster assignments with trip counts and justify the choice of K or test its robustness.
[§3.1.2, §5, Abstract] The comparative framing overstates what is actually analyzed. Section 3.1.2 states that the Pathao dataset has no geospatial coordinates and no datetime attribute, so geospatial analysis and time-series forecasting are performed only for the NYC taxi data. The Pathao contribution is limited to exploratory analysis of day-of-week and hour-of-day patterns. The introduction and abstract should be reframed to make this asymmetry explicit, rather than implying that both datasets support the full geospatial and temporal pipeline.

minor comments (6)

[Title] The title contains an apparent typo: 'A S TUDY' should be 'A STUDY'.
[§3.3] Section 3.3 says an 80:20 train-test split was performed on both datasets, but no model is evaluated on the Pathao data; clarify that the split applies only to the time-series analysis of the NYC taxi data.
[§4.3 and Figure 7] The text says the data are de-seasonalized, while Figure 7 is labeled 'De-Trended Time Series'; make the terminology consistent.
[§5.1] The phrase 'late hours (15th and 23rd hour)' is ambiguous; it should specify 15:00 and 23:00, and the claims about hourly peaks should be tied directly to the corresponding figures.
[References] Reference [9] appears in the reference list but is not cited in the body of the paper, and several references lack complete venue or pagination details; the bibliography needs cleanup.
[General] The paper does not include a data availability or reproducibility statement, and no code is provided; adding these would strengthen the contribution.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the SARIMAX forecast uses a train/test split and the weekly-seasonality claim is an empirical modeling choice, not a result forced by construction.

full rationale

The paper's central claims are descriptive data summaries and a SARIMAX forecast. The model order and seasonal order are selected from fitting diagnostics and the weekly periodicity is stated as a property of the data window, not derived from the target conclusion. The in-sample RMSE of 734.944 and out-of-sample RMSE of 204.1525 are reported as evaluation metrics after fitting, and the out-of-sample forecast uses a train/test split, so the prediction is not equivalent to the fitted input by construction. No equations define one result in terms of the claimed prediction, no fitted parameter is renamed as a forecast, and no self-citation is used as load-bearing evidence. The single-month data window and the short out-of-sample horizon raise legitimate statistical concerns about robustness, but those are validity issues, not circularity. Therefore the analysis is self-contained with no significant circular steps.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claims rest on the representativeness of one month of NYC data, the validity of the Pathao dataset as a Dhaka mobility proxy, and the unvalidated choices of SARIMAX orders and K=15. No new entities are introduced.

free parameters (3)

SARIMAX order (p,d,q) = (1,0,1)
Selected based on model fitting diagnostics (Section 5.3); no AIC/BIC or search procedure reported.
SARIMAX seasonal order (P,D,Q,s) = (1,0,1,7)
Chosen to account for weekly seasonality in one month of data (Section 5.3).
Number of K-means clusters K = 15
Set in Section 4.4 without justification or cluster validity metric.

assumptions (4)

domain assumption One month (January 2013) of NYC taxi data is sufficient to estimate and validate weekly seasonal demand patterns
The paper restricts to January for computational convenience (Section 3.1.1) and uses it to fit a weekly seasonal model (Section 4.3).
domain assumption The Pathao Food Trip dataset is an acceptable substitute for a Bangladeshi taxi dataset
Section 3 states this explicitly; the dataset has no coordinates or datetime, so it cannot support the geospatial or time-series analyses.
ad hoc to paper K-means with K=15 on raw latitude/longitude coordinates yields meaningful high/low demand regions
Section 4.4 applies K-means without preprocessing, distance weighting, or validation; cluster structure on raw coordinates may reflect geographic density rather than demand.
domain assumption SARIMAX residual assumptions (stationarity, no autocorrelation) hold after the detrending shown in Figure 7
Section 4.3 and 5.3 report only RMSE, not residual diagnostics or formal stationarity tests.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Geospatial and Temporal Trends in Urban Transportation: A Study of NYC Taxis and Pathao Food Deliveries." pith.science (2026). https://pith.science/paper/R7LBHOE6

@misc{pith2026250503816,
  author       = {Pith},
  title        = {Pith review of: Geospatial and Temporal Trends in Urban Transportation: A Study of NYC Taxis and Pathao Food Deliveries},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R7LBHOE6}},
  note         = {Machine review of arXiv:2505.03816}
}

read the original abstract

Urban transportation plays a vital role in modern city life, affecting how efficiently people and goods move around. This study analyzes transportation patterns using two datasets: the NYC Taxi Trip dataset from New York City and the Pathao Food Trip dataset from Dhaka, Bangladesh. Our goal is to identify key trends in demand, peak times, and important geographical hotspots. We start with Exploratory Data Analysis (EDA) to understand the basic characteristics of the datasets. Next, we perform geospatial analysis to map out high-demand and low-demand regions. We use the SARIMAX model for time series analysis to forecast demand patterns, capturing seasonal and weekly variations. Lastly, we apply clustering techniques to identify significant areas of high and low demand. Our findings provide valuable insights for optimizing fleet management and resource allocation in both passenger transport and food delivery services. These insights can help improve service efficiency, better meet customer needs, and enhance urban transportation systems in diverse urban environments.

Figures

Figures reproduced from arXiv: 2505.03816 by the authors.

**Figure 1.** Geospatial area in New York City In this study, we begin with Exploratory Data Analysis (EDA) to gain a basic understanding of the datasets and identify key trends. EDA involves visualizing and summarizing the data to uncover patterns, relationships, and anomalies. This helps us understand how taxi and food delivery services are used over time and across different locations. Next, we perform geospatial analysis to e… view at source ↗

**Figure 2.** Correlation Matrix 4 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** EDA on Pathao Food Trip Dataset 4.1.2 Pathao Food Trip Dataset To gain insights into the Pathao Food Trip Dataset, we performed the following analyses and their visual illustrations can be seen in [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: EDA on NYC Taxi Trip Dataset • Demand by Day of the Week (DoW): We analyzed the demand for food delivery services based on the day of the week to understand weekly demand patterns. • Demand by Hour of the Day (HoD): We examined the hourly demand for food deliveries to …

**Figure 6.** Figure 6: Geographical Clustering of NYC Taxi Trips [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

**Figure 7.** Figure 7: Detrending the Seasonal Data [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

**Figure 8.** Figure 8: Predicting the number of rides using SARIMAX model [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

PatiGonit22K: A Comprehensive Dataset for Solving Complex Bengali MWPs
cs.CL 2026-07 conditional novelty 4.0 of 10

PatiGonit22K is a new 22,441-problem Bengali math word problem dataset that adds 17,029 multi-operation complex problems to the existing PatiGonit benchmark.
GSM-Plus-BN: A Perturbation-Based Benchmark for Bangla Mathematical Reasoning in Large Language Models
cs.CL 2026-07 conditional novelty 4.0 of 10

The paper releases GSM-Plus-BN, a human-verified Bengali translation of the GSM-Plus perturbed math benchmark, and reports accuracy baselines for six open LLMs under standard and CoT prompting.

Reference graph

Works this paper leans on

15 extracted references · 14 canonical work pages · cited by 2 Pith papers

[1]

Analysis & Prediction of New York City Taxi and Uber Demands

Correa, Diego, and Christian Moyano. "Analysis & Prediction of New York City Taxi and Uber Demands." Journal of Applied Research and Technology 21.5 (2023): 886-898

work page 2023
[2]

Temporal geospatial analysis of secondary school students’ examination performance

Abd Kadir, ND Nik, and N. A. Adnan. "Temporal geospatial analysis of secondary school students’ examination performance." IOP Conference Series: Earth and Environmental Science. V ol. 37. No. 1. IOP Publishing, 2016

work page 2016
[3]

Revealing spatiotemporal travel demand and community structure characteristics with taxi trip data: A case study of New York City

Xie, Chen, et al. "Revealing spatiotemporal travel demand and community structure characteristics with taxi trip data: A case study of New York City." PLoS one 16.11 (2021): e0259694

work page 2021
[4]

Data Exploratory on Taxi Data in New York City

Archie, Christie Natashia, and Shubashini Rathina Velu. "Data Exploratory on Taxi Data in New York City." 2023 International Conference on Integrated Intelligence and Communication Systems (ICIICS). IEEE, 2023

work page 2023
[5]

ARIMA model for accurate time series stocks forecasting

Khan, Shakir, and Hela Alghulaiakh. "ARIMA model for accurate time series stocks forecasting." International Journal of Advanced Computer Science and Applications 11.7 (2020)

work page 2020
[6]

Time Series Forecasting using LSTM and ARIMA

Albeladi, Khulood, Bassam Zafar, and Ahmed Mueen. "Time Series Forecasting using LSTM and ARIMA." International Journal of Advanced Computer Science and Applications 14.1 (2023): 313-320

work page 2023
[7]

Profit prediction using ARIMA, SARIMA and LSTM models in time series forecasting: A comparison

Sirisha, Uppala Meena, Manjula C. Belavagi, and Girija Attigeri. "Profit prediction using ARIMA, SARIMA and LSTM models in time series forecasting: A comparison." IEEE Access 10 (2022): 124715-124727

2022
[8]

Forecasting Surabaya–Jakarta Train Passengers with SARIMA model

Astuti, S. W. "Forecasting Surabaya–Jakarta Train Passengers with SARIMA model." IOP Conference Series: Materials Science and Engineering. V ol. 407. No. 1. IOP Publishing, 2018

work page 2018

Show all 15 references

[9]

Optimization of Data Mining for Business Intelligence Using Cobweb Clustering Algorithm-A Case Study of Nigeria Network Service Providers’ Data Plan

Nnamani, Kelvin Ndubuisi, Oliver Okechukwu, and Chiamaka Chidimma Agu. "Optimization of Data Mining for Business Intelligence Using Cobweb Clustering Algorithm-A Case Study of Nigeria Network Service Providers’ Data Plan."
[10]

Unsupervised K-means clustering algorithm

Sinaga, Kristina P., and Miin-Shen Yang. "Unsupervised K-means clustering algorithm." IEEE access 8 (2020): 80716-80727. 10 A PREPRINT - M AY 8, 2025

2020
[11]

Geospatial analysis of the dynamics of climate in Kolkata metropolitan area

Talapatra, Adrita, Arijit Majumder, and Sandipan Das. "Geospatial analysis of the dynamics of climate in Kolkata metropolitan area." Journal of Physics: Conference Series. V ol. 1964. No. 4. IOP Publishing, 2021

1964
[12]

Revealing travel patterns and city structure with taxi trip data

Liu, Xi, et al. "Revealing travel patterns and city structure with taxi trip data." Journal of transport Geography 43 (2015): 78-90

2015
[13]

New York City taxi trip duration prediction using MLP and XGBoost

Poongodi, M., et al. "New York City taxi trip duration prediction using MLP and XGBoost." International Journal of System Assurance Engineering and Management (2022): 1-12

2022
[14]

The service operation of Foodpanda Bangladesh

Boron, Sami Islam. "The service operation of Foodpanda Bangladesh." (2022)

2022
[15]

Assessing the usability of ridesharing mobile applications in Bangladesh: an empirical study

Rahman, Md Mokhlesur, et al. "Assessing the usability of ridesharing mobile applications in Bangladesh: an empirical study." 2020 61st International Scientific Conference on Information Technology and Management Science of Riga Technical University (ITMS). IEEE, 2020. 11

2020

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

Analysis & Prediction of New York City Taxi and Uber Demands

Correa, Diego, and Christian Moyano. "Analysis & Prediction of New York City Taxi and Uber Demands." Journal of Applied Research and Technology 21.5 (2023): 886-898

work page 2023

[2] [2]

Temporal geospatial analysis of secondary school students’ examination performance

Abd Kadir, ND Nik, and N. A. Adnan. "Temporal geospatial analysis of secondary school students’ examination performance." IOP Conference Series: Earth and Environmental Science. V ol. 37. No. 1. IOP Publishing, 2016

work page 2016

[3] [3]

Revealing spatiotemporal travel demand and community structure characteristics with taxi trip data: A case study of New York City

Xie, Chen, et al. "Revealing spatiotemporal travel demand and community structure characteristics with taxi trip data: A case study of New York City." PLoS one 16.11 (2021): e0259694

work page 2021

[4] [4]

Data Exploratory on Taxi Data in New York City

Archie, Christie Natashia, and Shubashini Rathina Velu. "Data Exploratory on Taxi Data in New York City." 2023 International Conference on Integrated Intelligence and Communication Systems (ICIICS). IEEE, 2023

work page 2023

[5] [5]

ARIMA model for accurate time series stocks forecasting

Khan, Shakir, and Hela Alghulaiakh. "ARIMA model for accurate time series stocks forecasting." International Journal of Advanced Computer Science and Applications 11.7 (2020)

work page 2020

[6] [6]

Time Series Forecasting using LSTM and ARIMA

Albeladi, Khulood, Bassam Zafar, and Ahmed Mueen. "Time Series Forecasting using LSTM and ARIMA." International Journal of Advanced Computer Science and Applications 14.1 (2023): 313-320

work page 2023

[7] [7]

Profit prediction using ARIMA, SARIMA and LSTM models in time series forecasting: A comparison

Sirisha, Uppala Meena, Manjula C. Belavagi, and Girija Attigeri. "Profit prediction using ARIMA, SARIMA and LSTM models in time series forecasting: A comparison." IEEE Access 10 (2022): 124715-124727

2022

[8] [8]

Forecasting Surabaya–Jakarta Train Passengers with SARIMA model

Astuti, S. W. "Forecasting Surabaya–Jakarta Train Passengers with SARIMA model." IOP Conference Series: Materials Science and Engineering. V ol. 407. No. 1. IOP Publishing, 2018

work page 2018