Evaluating Federated Learning for Cross-Country Mood Inference from Smartphone Sensing Data

Sharmad Kalpande; Saurabh Shirke; Haroon R. Lone

cs.LGFeb 17, 2026

Evaluating Federated Learning for Cross-Country Mood Inference from Smartphone Sensing Data

Authors: Sharmad Kalpande, Saurabh Shirke, Haroon R. Lone

Organizations: Indian Institute of Science Education and Research Bhopal, India

Abstract

Mood instability is a key behavioral indicator of mental health, yet traditional assessments rely on infrequent and retrospective reports that fail to capture its continuous nature. Smartphone-based mobile sensing enables passive, in-the-wild mood inference from everyday behaviors; however, deploying such systems at scale remains challenging due to privacy constraints, uneven sensing availability, and substantial variability in behavioral patterns. In this work, we study mood inference using smartphone sensing data in a cross-country federated learning setting, where each country participates as an independent client while retaining local data. We introduce FedFAP, a feature-aware personalized federated framework designed to accommodate heterogeneous sensing modalities across regions. Evaluations across geographically and culturally diverse populations show that FedFAP achieves an AUROC of 0.744, outperforming both centralized approaches and existing personalized federated baselines. Beyond inference, our results offer design insights for mood-aware systems, demonstrating how population-aware personalization and privacy-preserving learning can enable scalable and mood-aware mobile sensing technologies.

Figures & tables

Country	Participants	Total Instances
CN	38	11523
DK	17	6424
IN	19	3071
MX	19	8400
PY	22	6467
UK	52	18832

Table 1. Participants and total instances (mood self-reports) across countries Participants and total instances (number of mood self-reports) across countries after preprocessing.

Sensor File	No. of Features	Extracted Features
activities	6	Fraction of time spent in: On Foot / In Vehicle / On Bicycle / Still / Tilting / Unknown activity
bluetooth	7	Number of unique Bluetooth device addresses; mean/min/max/std/var of RSSI; entropy of device addresses.
cellularnetwork	6	Number of unique cell towers; mean/min/max/std of signal strength; Entropy of cell-tower usage.
location	17	Centermost latitude and longitude (representing location); mean/min/max of latitude/longitude/altitude (capturing spatial coverage and elevation changes); mean/min/max/std of speed; Radius of gyration; Total distance traveled.
notification	2	Number of posted notifications; Number of removed notifications.
proximity	4	mean/std/min/max of proximity values.

Table 2. Summary of all sensor modalities and extracted features. Here, min, max, mean, std, and var denote the minimum, maximum, mean, standard deviation, and variance, respectively. RSSI refers to the Received Signal Strength Indicator.Entropy captures state diversity within a time window.

Figure 3. Distribution of feature missingness across countries, in which the marked areas denote the number of features belonging to different missingness ranges.

Figure 4. Proposed architecture comprising of a global server and N clients.

Case	5E 5R		5E 10R		10E 10R		10E 20R
	F1-w	AUROC	F1-w	AUROC	F1-w	AUROC	F1-w	AUROC
FedPer	0.663	0.512	0.663	0.516	0.667	0.520	0.665	0.518
pFedMe	0.662	0.494	0.662	0.512	0.662	0.518	0.662	0.491
FedFAP	0.686	0.553	0.689	0.577	0.693	0.591	0.704	0.627

Table 3. Comparison of baseline personalization methods with the proposed FedFAP under different training configurations (E: local epochs, R: communication rounds, F1-w: weighted F1 score).

Case	5E 10R		10E 10R		10E 20R
	F1-w	AUROC	F1-w	AUROC	F1-w	AUROC
Encoder	0.689	0.577	0.693	0.591	0.704	0.627
Attn + Encoder	0.689	0.576	0.691	0.592	0.706	0.637
1D-CNN	0.696	0.606	0.694	0.603	0.725	0.684

Table 4. Performance comparison of Encoder , Attention+Encoder , and 1D-CNN under different training configurations (E: local epochs, R: communication rounds, F1-w: weighted F1 score).

Setting	Averaging Method	F1-weighted	AUROC
	FedAvg	0.688	0.567
5E 5R	FedProx	0.689	0.563
	FedAdam	0.689	0.569
	FedAvg	0.696	0.606
5E 10R	FedProx	0.691	0.595
	FedAdam	0.693	0.593

Table 5. Performance comparison of FedFAP across different federated aggregation methods.

Figure 5. AUROC scores across different Federated averaging methods at 10 epochs and 50 rounds.

Figure 6. AUROC (y-axis) across countries (x-axis) for centralized baselines (XGBoost, RF, LR, 1D CNN) and the FedFAP.

Federated Learning

Federated learning trains shared models across data holders without requiring them to pool their raw data.

See all 943 papers

FedMental: Evaluating Federated Learning for Mental Health Detection from Social Media Data
May 18, 2026Nuredin Ali Abdelkadir, Anjali Ratnam, Zeerak Talat +1Depression Detection Mental Health
University of Minnesota · University of Edinburgh
Hardware-Aware Federated Learning for Speech Emotion Recognition
May 23, 2026Beyazit Bestami Yuksel, Emrah DikbiyikCommunication-Efficient Distributed Training Client Selection in Federated Learning
Computer Engineering Istanbul Technical University Istanbul, Turkey · Department of Computer Technologies Istanbul University-Cerrahpaşa Istanbul, Turkey
Uncertainty-Aware Federated Learning for Infant Movement Analysis
Sep 25, 2026Edmond S. L. HoFederated Learning Federated Learning Aggregation
School of Computing Science University of Glasgow Glasgow, Scotland, United Kingdom

May 18, 2026J/K move · Enter open · S save

FedMental: Evaluating Federated Learning for Mental Health Detection from Social Media Data

Nuredin Ali Abdelkadir, Anjali Ratnam, Zeerak Talat, Stevie Chancellor

University of Minnesota · University of Edinburgh

Depression Detection Mental Health Non-IID Federated Learning Privacy-Preserving ML Differentially Private Federated Learning Differential Privacy+2

Open paper AskPDF arXiv

Social media text data are often used to train Machine Learning (ML) models to identify users exhibiting high-risk mental health behaviors. However, sharing this sensitive data poses privacy risks and limits the growth of benchmark datasets. We comprehensively evaluate whether privacy-preserving ML techniques can enable safer data sharing while preserving performance. Specifically, we apply federated learning (FL) and Differentially Private FL for two widely-studied mental health prediction tasks: depression detection on X (Twitter) and suicide crisis detection on Reddit. We simulate realistic data-sharing scenarios by treating each user as a client in a non-IID setting, evaluating across different client fractions, aggregation strategies, and privacy budgets. While FL achieves comparable performance to centralized training (centralized F1 = 85.63; best FL model F1 = 83.16) on depression identification, we find that Differentially Private FL has a large performance-privacy trade-off (up to F1 = 27.01 drop) even with low levels of noise (epsilon = 50). This is due to the distortion of highly informative yet sparse mental health linguistic markers related to mental health, like health topics and emotion words. This research empirically demonstrates the potential and limitations of current privacy preservation techniques for mental health inference tasks.

Abstract

Federated Learning

Explore similar work

FedMental: Evaluating Federated Learning for Mental Health Detection from Social Media Data

Hardware-Aware Federated Learning for Speech Emotion Recognition

Uncertainty-Aware Federated Learning for Infant Movement Analysis