smc2021_c2
Solution for the SMC 2021 Data Challenge, Challenge 2 (Link Prediction)
The codebase assumes that the dataset (particularly, edges_cc.csv) is already stored in a ./data folder.
SMC.ipynb contains code that will convert edges_cc.csv to our evaluation dataset, consisting of data/modified_giant.csv (the giant component - 20% of the edges) and data/missing_edges.csv (the 20% missing edges)
louvain.py contains code that runs community detection using cugraph and networkx. Usage: python louvain.py ./data/modified_giant.csv 1.0
bc.py contains code that runs betweenness centrality using cugraph and networkx. Usage python bc.py ./data/modified_giant.csv
SMC-community.ipynb contains code to visualize the community structure obtained via Louvain.
SMC-cosine-similarity.ipynb contains code to create our predictive featuresets and evaluate simple threshold-based classifiers based off of those features. It also generates neural_network_training_data.csv, which contains a dataset with all 3 features.
networktraining_100epoch.ipynb contains the code for training the feedforward neural network. networktraining_100epoch_lowlr.ipynb does the same thing with a lower learning rate.