marthadais / TrajectoriesCompressionAnalysis

The proposed methodology assess how compression algorithms influence the clustering analysis with respect to anomaly detection of vessel trajectories.

TrajectoriesCompressionAnalysis

The proposed methodology assess how compression algorithms influence the clustering analysis with respect to anomaly detection of vessel trajectories. Results shows that a suitable compression algorithm for a particular scenario can reduce the overall processing time with a low impact on the clustering outcome. This source code is related to the work in [1]. Thus, if you are using this code please cite [1].

Files Description

1. vessel_analysis.py

   • Code to execute the analysis on vessels
   • Dataset can be found at https://figshare.com/s/3b736300bf47bffbcc07.
   • Select the vessel type:
     1. this line refers to fishing vessel

     # fishing vessels
     data_path = './data/crop/DCAIS_[30, 1001, 1002]_region_[37.6, 39, -122.9, -122.2]_01-04_to_30-06_trips.csv'
     

     2. this line refers to tankers vessel

     # tanker vessels
     data_path = './data/crop/DCAIS_[80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 1017, 1024]_region_[47.5, 49.3, -125.5, -122.5]_01-04_to_30-06_trips.csv'
     

   • Select the distance measure
     1. Define the metric as desired
     • 'dtw': dynamic time warping
     • 'hd': hausdorff distance
     • 'dfd': discrete fréchet distance
     • 'md': merge distance

     metric = 'dtw'
     

   • Select the minimum size of a cluster measure
     1. Define minimum size of a cluster measure
     • Fishing vessels: 2
     • Tanker vessels: 3

     msc = 2
     

2. source folder (src)

   • analysis.py: contains functions to analyze the different factors and plot images
     1. compression analysis
     2. distances analysis
     3. clustering analysis
   • clustering.py: contains clustering class to compute the clustering
   • compression.py: contains function to compute the compression
   • distance.py: contains functions to compute distance between trajectories

3. preprocessing folder

   • compress_trajectories.py: contains functions to read dataset and compute compression of all trajectories in the dataset

Requirements

The python version used was Python 3.9.5. The requirements to execute the code is in the file requirements.txt.

Reference

[1] Ferreira, M. D., Campbell, J. N., Purney, E., Soares, A., & Matwin, S. (2023). Assessing compression algorithms to improve the efficiency of clustering analysis on AIS vessel trajectories. International Journal of Geographical Information Science (TGIS)