Trajectories-Clustering

This project is realized as a computer science project in the Master 1 of data science at the university of Lille. 01/11/2020

Overview

This repository contains :

• trajectory_class.py : python script for the class Trajectory.

• grid_class.py : python script for the class Grid.

• test_trajectory.py : python script for testing the trajectory class with the TDD(Test-Driven Development) method.

• clustering trajectories.ipynb : A jupyter notebook version of the 3 sripts above that contains all the classes (Trajectory, Grid, Test)

Data

For the txt files used in this project, they are located in the folder cabspottingdata (1 file = 1 trajectory) which is available in this link.

Test-Driven Development (TDD) steps

• 1. Add a new unit test
• 2. Check it fails
• 3. Make it pass (with the others)
• 4. Refactor (if needed)
• 5. Commit (& push)
• 6. Go to (i).

TDD on Project

Identify input/output structures

• Input: Set of trajectories

  • Classes: Trajectory, Point

• Output: Graph of paths

  • Classes: Graph, Node, Edge

Apply TDD to develop clustering algorithm

Importing & plotting data

• Download the mobility dataset from here (SF cabs)

• Use the CSV python parser to load each filcabspottingdata.tar.gze in the directory (1 file = 1 trajectory) as an input trajectory for your algorithm

• Run clustering algorithm on the loaded dataset by creating a new Application class that combines the parser with the clustering algorithm

• Plot the resulting graph using matplotlib for basic rendering of the resulting graph, or:

• Explore the effects of changing the clustering “depths” on the resulting graph