<h2>About</h2><p>This series includes graph data science basics, mainly from <a href="https://neo4j.brand.live/c/2021nodes-homepage" target="_blank">neo4j NODES 2021</a> sessions.</p><ol><li>Part 1 : Graph data science Intro and high level data pipeline at neo4j environment</li><li>Part 2 : Basic Cypher Query</li><li>Part 3 : Bloom basics</li><li>Part 4 : Data Science Integration</li><li>Part 5 : Python/Jupyter integration</li></ol><p><br></p><h3>What is Graph Data Science (GDS)?</h3><p>Graph data science is an approach of doing data science using graph theory, gaining insights from the relationships and its structures.</p><blockquote>Graph data science techniques can be used as part of a variety of of different applications and use cases<ul><li>Graph queries support domain experts by answering common questions</li></ul><ul><li>Graph algorithms help make sense of the global structure of a graph, and the results used for standalone analysis or as features in a machine learning model</li></ul><ul><li>Graph embeddings are a core component of similarity graphs that power recommendation systems</li></ul><ul><li>Natural Language Processing techniques support content based filtering recommendations and knowledge graph completion</li></ul></blockquote><p>Ref : <a href="https://neo4j.com/developer/graph-data-science/" target="_blank">Neo4j Graph Data Science</a><br></p><p><br></p><p>Actual use cases on what Neo4j explains and what we could find on line like from medium are;</p><ul><li>Supply chain mapping/Bill of materials</li><li>Network management</li><li>Logistics and routing</li><li>Dependency tracing</li><li>Fraud detection</li><li>SNS</li></ul><p>Possible application would be impact study/similarity study in a certain market environment.</p><p><br></p><h4>Available libraries in neo4j</h4><p><img src="/media/django-summernote/2021-06-05/bc44d244-3293-4594-8bf6-c1392a4d37a6.png" style="width: 100%;"></p><p>This also includes good explanation for some of graph algorithms and its applications (ref : <a href="https://towardsdatascience.com/10-graph-algorithms-visually-explained-e57faa1336f3" target="_blank">10 Graph Algorithms Visually Explained</a>)</p><p><br></p><h4>Possible hypothetical applications for market analysis</h4><p>In the context of market environment analysis, followings would possibly provide some insights, and here are hypothetical expectations;</p><ol><li>Market centrality/influence/importance (PageRank, Eigenvector Centrality, Degree Centrality, etc) : To understand how much each company is influencing to the environment/market</li><li>Company Influencial Similarity (Node Similarity) : To understand which company behaves similarly in the market</li><li>Market community detection (Strongly connected components) : To identify community detection for a certain company</li><li>Link prediction/Filtering : To get a prescription on Supply chain vender/M&A recommendation</li></ol><p><br></p><h4>What is graph theory?</h4><p><img src="/media/django-summernote/2021-06-11/0943dc7e-0e3c-457f-a69e-1c23bc134504.png" style="width: 100%;"></p><p>Ref : <a href="https://commons.wikimedia.org/wiki/File:Social_Network_Analysis_Visualization.png" target="_blank">Social Network Analysis Visualization</a></p><p><br></p><blockquote><p>Graph theory is ultimately the study of relationships. Given a set of nodes & connections, which can abstract anything from city layout to computer data, graph theory provides a helpful tool to quantify & simplify the many moving parts of dynamic systems. Studying graphs through a framework provides answers to many arrangement, networking, optimization, matching and operational problems.</p></blockquote><p>Ref : <a href="https://towardsdatascience.com/what-is-graph-theory-and-why-should-you-care-28d6a715a5c2" target="_blank">What is Graph Theory, and why should you care?</a></p><h4><br></h4><h3 style="font-family: "Helvetica Neue", Helvetica, Arial, sans-serif; color: rgb(0, 0, 0);">What does neo4j do?</h3><blockquote>Neo4j is a native graph database platform, built from the ground up to leverage not only data but also data relationships. Neo4j connects data as it's stored, enabling queries never before imagines, at speeds never thought possible.</blockquote><p>Ref : <a href="https://neo4j.com/" target="_blank">What is Neo4j?</a></p><p><br></p><h4>Neo4j environment;</h4><p><img src="/media/django-summernote/2021-06-11/12203289-5990-4b21-bae9-c6dc0f9d8793.png" style="width: 100%;"></p><p>Cypher is open source query language to be used for neo4j graph database for store and retrieve data. GraphQL can be also used with the combination of javascript. Data is from various types of storage, SQL, NoSQL, CSV/JSON. Any data science result can be also added to graph database for a further analysis using graph database/datascience. The data is then combined into single place by neo4j database, either Aura, Sandbox, or local desktop neo4j application. These are data preparation pipeline for graph database.</p><p>Aura is cloud database offered by neo4j where they offer free tier plan (still as trial period though), while Sandbox is also cloud database but it is only used as a temporal server and data will be deleted in few days.</p><p>Graph database can be then drilled down at query browser to get an insight on relationship and context between data. The browser query also offers graph datascience analysis capability which provide various types of data science technique written above as GDS library.</p><p>Graph database also can be retrieved by online visualization application called `bloom`, which gives user BI like capability with their intuition. </p>
<< Back to Blog Posts
Back to Home