Title

FedAPT: Inter-Floor Noise Complaint Prediction in Residential Complexes via Federated Learning

Alternative Title

FedAPT: 연합학습 기반 공동주택 층간소음 민원 예측

DGIST Authors

Nakheon Ko ; Yongseob Lim ; Sungpil Chun

Advisor

임용섭

Co-Advisor(s)

Sungpil Chun

Issued Date

2025

Awarded Date

2025-08-01

Type

Thesis

Description

Federated Learning, Inter-Floor Noise Prediction, GRU, Neural Prophet, Smart Residential Systems

Table Of Contents

Ⅰ. BACKGROUND 1
1.1. Motivation 1
1.2. Previous Works 3
1.3. Original Contributions 4
1.4. Organization of the Study 5
Ⅱ. RELATED WORKS 6
2.1. Prior Research on Inter-Floor Noise Issues 6
2.2. Multivariate Time-Series Analysis Techniques for Forecasting Applications 8
2.3. Application of Federated Learning for Privacy-Preserving Prediction Tasks 9
Ⅲ. MATERIALS AND METHODS 11
3.1. Overall Architecture of the Proposed Forecasting System 11
3.2. Data Collection and Preprocessing 12
3.3. Multivariate Time-Series Forecasting Models 14
3.4. Implementation Plan for Federated Learning with FedAvg Aggregation 23
Ⅳ. EVALUATIONS AND RESULTS 25
4.1. Comparative Analysis of Model Performance 25
4.2. Performance of Models in Federated Learning Settings 34
Ⅴ. DISCUSSIONS AND CONCLUSIONS 40
5.1. Conclusions 40
5.2. Limitations and Future Works 41
References 44
국문 초록 48

URI

https://scholar.dgist.ac.kr/handle/20.500.11750/59821
http://dgist.dcollection.net/common/orgView/200000888731

DOI

10.22677/THESIS.200000888731

Degree

Master

Department

Department of Robotics and Mechatronics Engineering

Publisher

DGIST

Show Full Item Record