The HOG-FDA Approach with Mobile Phone Data to Modeling the Dynamic of People's Presences in the City

Maurizio Carpita; Rodolfo Metulini

arxiv: 1907.05026 · v1 · pith:WX7MAYEYnew · submitted 2019-07-11 · 📊 stat.AP

The HOG-FDA Approach with Mobile Phone Data to Modeling the Dynamic of People's Presences in the City

Rodolfo Metulini , Maurizio Carpita This is my paper

Pith reviewed 2026-05-24 23:07 UTC · model grok-4.3

classification 📊 stat.AP

keywords mobile phone datasmart cityHOGfunctional data analysisspatio-temporal modelingpeople presencemodel-based clusteringBrescia

0 comments

The pith

Mobile phone data analyzed with HOG for space and FDA for time shows city presence follows consistent seasonal and weekday patterns.

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

The paper establishes that high-dimensional spatio-temporal mobile phone data can be processed by first extracting spatial structure via the Histogram of Oriented Gradients method and then applying model-based clustering on functional data to identify temporal profiles. This reveals regular similarities among days that align with seasons or days of the week. A reader would care because the resulting models quantify how the number of people in a city varies between 30 and 60 thousand depending on season, weekday, and time of day, supporting better planning and anomaly detection in smart-city contexts. The approach is demonstrated on data from the Municipality of Brescia.

Core claim

The HOG-FDA approach, combining Histogram of Oriented Gradients to capture spatial structure with Model-Based Clustering Functional Data Analysis to capture temporal evolution, applied to mobile phone data in Brescia identifies similarities among days that follow a seasonal or days-of-the-week trend, with the number of users in the city varying from 30 to 60 thousand depending on the season, the day of the week and the time of the day.

What carries the argument

The HOG-FDA pipeline: Histogram of Oriented Gradients extracts oriented spatial features from the high-dimensional data grid, while Model-Based Clustering Functional Data Analysis groups the resulting daily temporal curves into clusters that reflect recurring patterns.

If this is right

Daily profiles can be grouped into a small number of clusters that correspond to expected seasonal and weekday categories.
Deviations from these clusters can serve as indicators of anomalies or special events.
City services such as transport and utilities can be scaled according to the identified 30-60 thousand user range tied to specific times and days.
The same pipeline can be rerun on updated data streams to track whether the observed seasonal and weekly regularities persist over longer periods.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Extending the method to live data feeds could support real-time capacity alerts for public spaces.
Similar processing of phone data from other cities would test whether the seasonal-weekday structure generalizes beyond Brescia.
Combining the output clusters with external variables such as weather or events could improve predictive models for crowd levels.

Load-bearing premise

Mobile phone records give a representative sample of actual human presence without major biases from coverage gaps, user demographics, or data collection artifacts.

What would settle it

Direct comparison against independent pedestrian counts or traffic sensors in Brescia that shows no seasonal or weekday clustering in presence levels, or variation outside the 30-60 thousand range, would falsify the claimed regularities.

read the original abstract

In the context of Smart City, the dynamic of the presence of people can be analysed using high-dimensional spatio-temporal mobile phone data. In order to find regularities and detect anomalies in the daily profiles, we propose an approach that considers the spatial structure by means of Histogram of Oriented Gradients (HOG) method and the temporal evolution using a Model-Based Clustering Functional Data Analysis (FDA). An application to the case study of the Municipality of Brescia is provided. Similarities among days, that follow a seasonal or a days of the week trend, exist. The number of users in the city, depending on the season, the day of the week and the time of the day, varies from 30 to 60 thousands of people

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Straight application of HOG plus FDA clustering to Brescia mobile data that shows seasonal and weekly patterns but supplies no validation or bias checks.

read the letter

The main takeaway is that this paper takes Histogram of Oriented Gradients to encode spatial structure in mobile phone counts and then applies model-based functional data analysis clustering to the resulting daily profiles. They run it on data from Brescia and report that days group by season or weekday, with total users ranging 30-60k depending on time and day type. That specific pairing on this data source is the only real novelty; the methods themselves are standard. The case study is laid out clearly enough to show how the workflow produces those clusters. The soft spots are the absence of any reported validation steps, baseline comparisons, error measures, or checks on data quality. The abstract gives no evidence that the phone counts represent the full population rather than one carrier's users or a skewed demographic slice, so the clusters could reflect sampling artifacts instead of actual presence. The stress-test concern about representativeness lands directly on what is shown. This is for urban analytics practitioners who need a worked example of these tools on spatio-temporal phone data. A reader looking for ideas on implementation might find it useful, but anyone wanting substantiated claims will see the gaps. It deserves peer review so referees can ask for the missing calibration and robustness checks; the application itself is narrow enough that it does not need to be groundbreaking to be worth referee time.

Referee Report

2 major / 2 minor

Summary. The paper claims that the HOG-FDA approach, combining Histogram of Oriented Gradients for spatial structure and Model-Based Clustering Functional Data Analysis for temporal evolution, can model the dynamic of people's presences using mobile phone data. In the Brescia case study, it reveals similarities among days following seasonal or day-of-week trends, with the number of users varying from 30 to 60 thousand depending on season, day, and time.

Significance. If the findings hold after validation, the work shows a cross-disciplinary application of HOG (from computer vision) and FDA clustering to high-dimensional telecom data for extracting urban presence regularities, potentially useful for smart-city analytics. No machine-checked proofs, reproducible code, or parameter-free derivations are reported.

major comments (2)

[Abstract] Abstract: the stated results on seasonal/day-of-week clusters and the 30-60k user range are presented with no validation details, error bars, data quality checks, or comparison to baselines or ground truth, so the central claim that HOG-FDA models actual presence dynamics rests on unshown analysis.
[Case study of the Municipality of Brescia] Case study / results: the claim that mobile-phone counts accurately proxy real human presence (required to interpret the HOG spatial features and FDA temporal clusters) is not supported by any calibration, demographic-bias check, or coverage analysis, which is load-bearing for the reported patterns.

minor comments (2)

[Abstract] The abstract contains minor grammatical issues (e.g., 'the dynamic of the presence of people').
[Methods] Notation for the FDA model-based clustering step could be made more explicit to aid reproducibility.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the detailed comments. We respond point by point below, indicating where revisions will be made to address concerns about validation and proxy assumptions.

read point-by-point responses

Referee: [Abstract] Abstract: the stated results on seasonal/day-of-week clusters and the 30-60k user range are presented with no validation details, error bars, data quality checks, or comparison to baselines or ground truth, so the central claim that HOG-FDA models actual presence dynamics rests on unshown analysis.

Authors: The abstract condenses the main outcomes observed in the Brescia case study. The manuscript body describes the HOG feature extraction and FDA clustering steps applied directly to the mobile-phone counts. We agree that the abstract does not convey the exploratory character of the work or the absence of external validation. We will revise the abstract to state that the reported patterns are derived from the given dataset without ground-truth comparisons, and we will add a limitations paragraph in the discussion section. revision: partial
Referee: [Case study of the Municipality of Brescia] Case study / results: the claim that mobile-phone counts accurately proxy real human presence (required to interpret the HOG spatial features and FDA temporal clusters) is not supported by any calibration, demographic-bias check, or coverage analysis, which is load-bearing for the reported patterns.

Authors: The paper treats mobile-phone counts as a standard proxy for presence, consistent with prior telecom-data literature. No calibration data, demographic surveys, or coverage statistics were available for the Brescia dataset used here. We will add an explicit subsection acknowledging this assumption, its potential biases (e.g., age or socioeconomic skew), and the consequent limits on interpreting the spatial HOG features and temporal clusters as direct measures of population presence. revision: yes

standing simulated objections not resolved

Provision of external validation, error bars, baseline comparisons, or ground-truth calibration for the mobile-phone proxy, as no such auxiliary data existed for the study.

Circularity Check

0 steps flagged

No circularity: application of HOG and FDA to mobile data with no self-referential derivation

full rationale

The paper applies established Histogram of Oriented Gradients (HOG) for spatial structure and Model-Based Clustering Functional Data Analysis (FDA) for temporal evolution to mobile phone data from Brescia. The abstract and description present this as a methodological application to detect seasonal and day-of-week patterns in user counts (30-60k range), without any derivation chain, parameter fitting presented as prediction, or self-citation that reduces the central claim to its own inputs. No equations or steps are quoted that exhibit self-definition, fitted-input renaming, or uniqueness imported from prior author work. The analysis is self-contained as an empirical case study; representativeness of the data is an external assumption, not a circularity issue.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are stated or derivable from the provided text.

pith-pipeline@v0.9.0 · 5654 in / 991 out tokens · 20237 ms · 2026-05-24T23:07:05.882127+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We propose an approach that considers the spatial structure by means of Histogram of Oriented Gradients (HOG) method and the temporal evolution using a Model-Based Clustering Functional Data Analysis (FDA).
IndisputableMonolith/Foundation/DimensionForcing.lean alexander_duality_circle_linking unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Similarities among days, that follow a seasonal or a days of the week trend, exist.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.