Source code for the masters thesis "Finding Candidate Node Pairs for Link Prediction at Scale" by Kalkan, Filip and Hambiralovic, Mahir.

This repository includes a benchmarking framework for testing candidate selection algorithms, along with implementations of some candidate selection algorithms, one of them being DAPPR.

A number of of datasets of varying domains and sizes are available for testing.

The project requires Python 3.10.

Using conda, install all dependencies in environment.yml.

Once dependencies are installed, you can try running the benchmarks in multi.ipynb.

Tools for vizualizing the results are available in src/graphs.ipynb.

DAPPR can be run in src/main.py. Example:

python main.py --edgelist-path "dataset/static/yeast.txt" --c 1000 --lambd 40 --parallel y -out candidate_node_pairs.csv

This example runs the YEAST dataset and outputs the candidate node pairs to a csv file named candidate_node_pairs.csv.