Sample applications using Parallel Graph AnalytiX (PGX)
Table of contents:
- Overview
- Download PGX
- Healthcare Fraud Detection
- Super Hero Network Analysis
- Graph-based ML applications using PgxML
- Article Ranking
- Movie Recommendation
- Entity Linking
- Research Paper Classification
Overview
This repository contains a set of examples and use cases that illustrate the capabilities of PGX. Some of these use cases act as examples for some advanced functionalities, such as:
- The PgxML library, for Graph-based ML
- The PGX Algorithm API, a high-level DSL for developing optimized graph algorithms.
Download PGX
PGX is included as part of the Oracle Graph Server and Client product.
A trial of Oracle Graph Server and Client can be downloaded from the Oracle Technology Network.
PgxML and PGX Algorithm are available as of version 3.2.0 and released under the OTN license.
Obtain the latest pgx-x.y.z-server
zip file from the Oracle Graph Server and Client Downloads page and unzip it in the libs
folder.
Healthcare Fraud Detection
The healthcare fraud detection example detects anomalies in medical transactions through a graph analysis procedure implemented in PGX. More details regarding this use-case are available here.
Super Hero Network Analysis
The Super Hero Network Analysis example describes how to combine computational graph analysis and graph pattern matching with PGX. More details regarding this use-case are available here.
Graph-based ML applications using PgxML
We provide two Graph-based ML applications, namely, Graphlet representation
and Node representation
.
Graphlet representation
This application demostrates how we can extract vector representation for each graphlet in a cluster of graphlets. For this application, we use the PG2Vec algorithm. More details regarding this application are available here.
Node representation
This application demonstrates how we can extract vector representation for each node in a graph. For this application, we use the DeepWalk algorithm. More details regarding this application are available here.
Article Ranking
This application demonstrates how ArticleRank could be employed to measure the influence of journal articles. More details regarding this application are available here.
Movie Recommendation
This application demonstrates how Matrix Factorization could be employed to recommend movies to users. More details regarding this application are available here.
Entity Linking
Entity Linking allows to connect Named Entities (for example, names of famous people) to their Wikipedia/DBpedia page. This application leverages vertex embeddings to provide high-quality results. More details available here and in our paper.
Research Paper Classification
This application demonstrates how graph data can be used to enhance classification performance of a research paper classifier. More details regarding this application are available here.