Comparative-Analysis-of-Traditional-Methods-and-Graph-Machine-Learning-Methods-for-Link-Prediction

Comparative Analysis of Traditional Methods and Graph Machine Learning Methods for Link Prediction on Game of Thrones Season 3 data
To compare traditional, similarity-based methods with GraphML algorithms like GCN, GraphSAGE and GAT for link prediction.

Perform EDA on the dataset
- Check for isolates, self-loops, etc
Calculate the following measures:
- Betweenness
- PageRank
- Local clustering coefficient
Find communities using Spectral Clustering
Link Prediction using Traditional Methods
- Perform comparison for 3 similarity measures Jaccard, Adamic Adar and Preferential Attachment.
- Finding the best similarity measure along with the optimal threshold for said similarity measure.
Link Prediction using GraphML
- Perform link prediction using
  - GCN
  - GraphSAGE
  - GAT
- Comparing the three models using Loss and AUC.
Performing a comparison between traditional and GraphML
- What are the metrics?
- Which is performing better? Is there any reason you can think of, as to why this might be happening?
- Any analysis or insights you can draw from this, that may relate to the season’s plot?

Shamanthkolegodu/Comparative-Analysis-of-Traditional-Methods-and-Graph-Machine-Learning-Methods-for-Link-Prediction