lesteraleong/crunchbasegraph

IPO prediction using CrunchBase Graph on TigerGraph Cloud

IPO Prediction in With Graph Convolutional Neural Network

This repository is a demo to show the process of forming a data science project around data that is stored in a TigerGraph graph database. It uses Giraffle with Gradle to install GSQL queries onto the database instance, and then call those REST endpoints with pyTigerGraph, a Python wrapper written to make accessing TigerGraph databases easy. We then feed the data that we recieve from the database into a Graph Convolutional Neural Network. Due to various simplifications and computational constraints, the GCN doesn't perform that great in classifying which companies will IPO or not, but it does serve as a good demo of what can be done with other datasets.

Setup TigerGraph Cloud Instance

To setup your own version of the server, head over to tgcloud.io and create a free account. It will walk you through creating your own cloud database instance, but make sure to choose the CrunchBase knowledge graph as the template to get started with.

Getting Gradle Installed and Setup

Follow the directions here to get Gradle installed on your machine. Once that is done, you will need to get an SSL certification from your cloud instance. To do this, run:

openssl s_client -connect hostname.i.tgcloud.io:14240 < /dev/null 2> /dev/null | \
openssl x509 -text > cert.txt

Installing the Python Packages Needed

You need to install quite a few Python packages. Good news, is that this is easily done via pip:

pip3 install pytigergraph
pip3 install torch
pip3 install dgl
pip3 install networkx

You are now all ready to try out the notebook. Find it in /py_scripts.

Credits

Demo/tutorial written by Parker Erickson, a student at the University of Minnesota pursuing a B.S. in Computer Science. His interests include graph databases, machine learning, travelling, playing the saxophone, and watching Minnesota Twins baseball. Feel free to reach out! Find him on:

• LinkedIn: https://www.linkedin.com/in/parker-erickson/
• GitHub: https://github.com/parkererickson
• Medium: https://medium.com/@parker.erickson
• Email: [email protected]

---

GCN Resources:

• DGL Documentation: https://docs.dgl.ai/
• GCN paper by Kipf and Welling https://arxiv.org/abs/1609.02907
• R-GCN paper: https://arxiv.org/abs/1703.06103

---

Notebook adapted from: https://docs.dgl.ai/en/latest/tutorials/basics/1_first.html