Shall I work with them? A ‘knowledge graph’-based approach for predicting future research collaborations

This repository hosts code for the paper: Shall I work with them? A ‘knowledge graph’-based approach for predicting future research collaborations

Installation

Prequisites:

• Python 3.6
• Neo4j (version 4.1.8 or greater).

Setup

git clone https://github.com/nkanak/cordkel
cd cordkel
pip install -r requirements.txt

Datasets

Available here

Experiments

For our experiments we have already generated the datasets found in the link above, so you only need to run the predict_all.py script.

Data Generation

To generate the data for our experiements, we firstly create the database from the Neo4j desktop app using 4.1.8 as the min. version.
Then we set the dbms.memory.heap.max_size = 4G in the database settings and we install the Graph Data Science plugin (version 1.4.1). We run the CORD19_GraphOfDocs.py script which creates thousands of nodes, and millions of relationships in the database.
Once it is done, the database is initialized and ready for use. We generate the base datasets, by running generate_datasets.py. Finally, we produce all similarity features for each generated .csv by running each .py script postfixed with _all.

Contributors

• Nikos Kanakaris (nkanakaris@upnet.gr)
• Nikolaos Giarelis (giarelis@ceid.upatras.gr)
• Ilias Siachos (ilias.siachos@upnet.gr)
• Nikos Karacapilidis (karacap@upatras.gr)