Federated-Learning-Based-Similarity-Search-over-Trajectory-Data-Federation

Requirments

• CentOS Linux release 7.9.2009 (Core)
• pip install -r reqirments.txt
• dataset: the Beijing Taxi Dataset

Dataset

• For privacy reasons, we are unable to provide the Beijing taxi dataset. If you wish to use the Porto dataset, you can download it from Porto Taxi Trajectory Data.

Data processing

Porto

• use bash scripts/data_processing.sh

Beijing

1. use utils/beijing2traj.py to convert the original trajectory points of the Beijing dataset into trajectories.
2. save the file obtained in the first step to data/beijing.csv.
3. use utils/preprocessing_beijing.py to process the data.

Run

If you are using beijing dataset, you may modify all 'dataset' field in scripts to 'beijing'.

ourmethod

• distort
  • use bash scripts/run_fed-trajCl_distort.sh
• downsampling
  • use bash scripts/run_fed-trajCl_downsampling.sh
• hit_ratio
  • use bash scripts/run_fed-trajCl_simi.sh

fedavg

• distort
  • use bash scripts/run_fedavg_distort.sh
• downsampling
  • use bash scripts/run_fedavg_downsampling.sh
• hit_ratio
  • use bash scripts/run_fedavg_simi.sh