Rough DBSCAN Implementation

An implementation of the paper
Rough-DBSCAN: A fast hybrid density based clustering method for large data sets
by P. Viswanath and V. Suresh Babub

Author

Marina Bermúdez Granados - nara-on

Project Structure

  .
  ├── datasets                       # Directory containing the different datsets
  ├── experiments                    # Directory containing all the main experiments
  ├── models                         # Directory containing the models: Counted-Leaders, DBSCAN, and Rough-DBSCAN
  ├── tests                          # Directory containing all the tests, single runs of DBSCAN and Rough-DBSCAN
  ├── utils                          # Directory containing different functionalities 
  └── visuals
      ├── experiments                # Directory for the results of the experiments
      └── tests                      # Directory for the results of the tests

Images

Classification of Data Points in Rough-DBSCAN

Some Rough-DBSCAN Results in Blobs Dataset with Different Parameters

Some Rough-DBSCAN Results in Banana Dataset with Different Parameters

Version History

• Updates to README.md
  • Improvements for GitHub
• Requirements
  • Update to requirements.txt
• Correction
  • Fixed typo
• More details
  • Extra experiments for PenDigits dataset
• Details
  • Minor optimizations

• Scratch Experiments
  • Uploaded scratch results datasets for Banana, Blobs, Letter, PendDigits, and Shuttle datasets

• Sklearn Experiments + Fixing issues
  • Added sklearn code to experiment files
  • Uploaded sklearn results datasets for Banana, Blobs, Letter, and Shuttle datasets
  • Minor corrections to timelimits in DBSCAN
• Experiments PenDigits + Optimizations
  • Developed experiment code for Banana dataset
  • Renamed DBSCAN class to DBSCAN_scratch
  • Updated testing code
  • Uploaded results dataset for leaders
  • Uploaded sklearn results dataset for PenDigits dataset
• Experiment Leaders vs Patterns
  • Updated hyperparameters in experiments
  • Added default parameters to synthetic dataset
  • Developed plotting functions for synthetic datasets
  • Uploaded visuals for Banana and Blobs dataset
  • Uploaded visuals for Leaders experiment

• Checkpoint Experiments
  • Tests folder created
  • Moved testing code to tests folder
  • Developed experiment code for blobs dataset
  • Developed experiments code to test the Leaders
  • Added timelimit to DBSCAN
  • Moved plotting resources to separate file in utils
  • Moved previous results to a separate tests folder in visuals
  • Uploaded PenDigits results in a separate experiments folder in visuals
  • Uploaded dense points documentation to visuals

• Testing Done
  • Updated PenDigits dataset
  • Developed experiment code for PenDigits
  • Updating hyperparameters for rest of experiments
  • Reorganized fit and predict of DBSCAN
  • Optimized find_N function
  • Reorganization of visuals folder with new content

• Checkpoint: Real datasets tests
  • Updates to Shuttle dataset
  • Developed experiment code for Letter and SHuttle datasets
  • Minor printing modifications to DBSCAN and Rough_DBSCAN
  • Minor updates to plot generation and dataset loading util functions
  • Upload first Letter and Shuttle experiments
• Fixed N, RDBSCAN Synthetic Tests
  • Developed find_rough_N utils function
  • Updated find_N utils funtcion
  • Updated Counted_Leaders class
  • Updates to DBSCAN for dual implementation of DBSCAN and Rough_DBSCAN
  • Updates to Rough_DBSCAN
  • Updates to plotting util functions
  • Renamed experiment codes to tests
  • Uploaded first set of visual results with different parameters
• Refactor Experiments
  • Developed code for the experiments of the remianing dataset
  • Developed dataset generation utils function for each dataset
  • Developed experiments utils function
  • Developed plotting utils function
  • Updated old results

• Mini Refactor Experiments
  • Datasets folder created
  • Upload PenDigits, Letter, and Shuttle datasets to datasets folder
  • Update DBSCAN for fit_predict function
  • Update Rough_DBSCAN for fit_predict function
  • Developed generate_results utils function
  • Uploaded and refactored visuals of banana and blobs datasets

• Initial fit Rough DBSCAN
  • Experiments folder created
  • Visuals folder created
  • Updates to Counted_Leaders class
  • Updates to Rough_DBSCAN class
  • Developed separate DBSCAN class
  • Developed find_N utils function
  • Developed code to experiments in banana and blobs dataset
  • Uploaded first results of experiments in banana and blob dataset

• Counted Leaders
  • Updates to Counted_Leaders class
  • Updates in coincidence function
• Developing Counted Leaders
  • First version of Counted Leaders
  • Developed coincidence utils function
  • Update README.md
• Basic structure
  • Models folder created
  • Utils folder created
  • Base code for Counted_Leaders and Rough_DBSCAN classes
  • First version of README.md
  • First version of requirements.txt
  • Added gitignore file
• Initial Commit
  • Initial Release