ravnoor-kaur/traffic-fl

How ravnoor-kaur/traffic-fl works

This research project compares hourly traffic forecasting with a last-hour baseline, centralized GRU training, and Flower-based federated GRU training across four junctions. The workflow runs from traffic data exploration and per-junction preprocessing through model training and test evaluation, then records and compares results. Federated clients train locally and exchange model weights with a FedAvg server. The graph preserves the documented exploration and experiment branches; relationships for unsampled modules are included only where the README or source index supports them.

Main entry points: server_app.py FedAvg server · train_central.py Central training · client_app.py Junction client · compare.py Result comparison

Components 14

Data preparation

Data exploration src/explore.py
Junction preprocessing src/data_prep.py

Training experiments

GRU model and training src/model.py
Central training src/train_central.py

Federated learning

Federated task helpers federated/trafficfl/task.py
Junction client federated/trafficfl/client_app.py
FedAvg server federated/trafficfl/server_app.py

Evaluation and results

Experiment log src/model.py
Round metrics federated/trafficfl/server_app.py
Result comparison src/compare.py
Result artifacts results

Other components

Researcher
Traffic CSV
Flower runtime

Connections 24

Researcher → Data exploration: runs exploration README.md
Traffic CSV → Data exploration: reads data src/explore.py
Researcher → Junction preprocessing: runs preprocessing README.md
Traffic CSV → Junction preprocessing: loads and sorts src/data_prep.py
Researcher → Central training: runs experiment README.md
Central training → Junction preprocessing: prepares junctions src/train_central.py
Central training → GRU model and training: trains and evaluates src/train_central.py
Central training → Experiment log: records metrics src/train_central.py
Researcher → FedAvg server: starts federation README.md
FedAvg server → Federated task helpers: loads tools and data federated/trafficfl/server_app.py
Federated task helpers → Junction preprocessing: prepares junction data federated/trafficfl/task.py
Federated task helpers → GRU model and training: provides model tools federated/trafficfl/task.py
FedAvg server → Flower runtime: starts FedAvg federated/trafficfl/server_app.py
Flower runtime → Junction client: dispatches rounds README.md
Junction client → Federated task helpers: loads local data federated/trafficfl/client_app.py
Junction client → GRU model and training: trains locally federated/trafficfl/client_app.py
Junction client → Flower runtime: returns weights federated/trafficfl/client_app.py
FedAvg server → Round metrics: writes round metrics federated/trafficfl/server_app.py
FedAvg server → Experiment log: logs final metrics federated/trafficfl/server_app.py
Researcher → Result comparison: runs comparison README.md
Result comparison → Experiment log: reads results src/compare.py
Result comparison → Round metrics: reads rounds src/compare.py
Result comparison → Result artifacts: writes tables and charts src/compare.py
Data exploration → Result artifacts: saves exploration charts src/explore.py

Drawn by GitDiagram on November 10, 2026 from the repository's file tree, README and source files. Read as Markdown ·