Neo4j Announces First Graph Machine Learning for the Enterprise

SAN MATEO, Calif., Oct. 20, 2020 /PRNewswire/ — Neo4j, the chief in graph know-how, introduced the most recent model of Neo4j for Graph Data Science, a breakthrough that democratizes superior graph-based machine studying (ML) strategies by leveraging deep studying and graph convolutional neural networks.

Until now, few firms outdoors of Google and Facebook have had the AI foresight and sources to leverage graph embeddings. This highly effective and modern approach calculates the form of the encompassing community for every bit of information within a graph, enabling much better machine studying predictions. Neo4j for Graph Data Science model 1.four democratizes these improvements to upend the way in which enterprises make predictions in various eventualities from fraud detection to monitoring buyer or affected person journey, to drug discovery and information graph completion. 

Neo4j for Graph Data Science model 1.four is the primary and solely graph-native machine studying performance commercially out there for enterprises. The capacity to study generalized, predictive options from information is critical as a result of organizations do not at all times know symbolize linked information to be used in machine studying fashions. The newest Neo4j model consists of graph embedding algorithms that study the construction of a consumer’s graph, somewhat than counting on predetermined formulation to calculate particular options like centrality scores. 

Alicia Frame, Neo4j’s Lead Product Manager and Data Scientist, shared what Neo4j for Graph Data Science model 1.four means for information scientists and analytics groups.

“We are thrilled to bring cutting-edge graph embedding techniques into easy-to-use enterprise software,” Dr. Frame stated. “The latest version of Neo4j for Graph Data Science democratizes state-of-the-science techniques and makes it possible for anyone to use graph machine learning. This is a game changer in terms of what can be achieved with predictive analysis.”

Graph Embeddings at UK.GOV

In GOV.UK’s latest weblog submit titled “One Graph to rule them all“, information scientists Felisia Loukou, and Dr. Matthew Gregory write about deploying their first machine studying mannequin with the assistance of graph information science and a Neo4j information graph. Their mannequin mechanically recommends content material to GOV.UK customers primarily based upon the web page they’re visiting. In their August 2020 submit, they clarify:

“node2vec, given any graph, learns continuous feature representations (a vector of numbers) for the nodes, which can then be used for various machine learning tasks, such as recommending content. Through this process, we learned that creating the necessary data infrastructure which underpins the training and deployment of a model is the most time-consuming part.”

Key Features

With Neo4j for Graph Data Science, organizations now have a totally new option to study from their information, get extra worth out of current datasets and regularly enhance predictive accuracy:

  • Uncover revelations of their information that they did not even know to search for: Graph embedding (algorithms) study what’s structurally vital in information assembling a generalized superset of data that every one the normal graph algorithms collect. Graph embeddings accomplish this by sampling the topology and properties of the graph after which lowering its complexity to simply these vital options for additional machine studying.
  • Eliminate plateaus when conventional algorithms aren’t sufficient. Graph algorithms and embeddings can summary the construction of a graph utilizing its topology and properties, making it potential to foretell outcomes primarily based on the connections between information factors – somewhat than uncooked information alone.
  • Perform quicker function engineering on information utilizing generalized studying to keep away from testing a mess of focused algorithms when predictive options are ambiguous and utilizing high-performance strategies like FastRP.
  • Continually incorporate new information and predictions by storing the discovered features of GraphSage within the new machine-learning mannequin catalog and making use of them to new information for brand new embeddings and predictions – with out having to retrain your mannequin.
  • Enhance the worth of their graph database by including ongoing scoring and classification outcomes, in addition to predicting lacking info for regularly higher insights.

Neo4j for Graph Data Science model 1.four consists of three new graph embedding choices that study graph topology to calculate extra correct representations:

  • node2Vec is a widely known graph embedding algorithm which makes use of neural networks
  • FastRP is a graph embedding as much as 75,000 instances quicker than node2Vec, whereas offering equal accuracy and scaling nicely even for very giant graphs
  • GraphSAGE is an embedding algorithm and course of for inductive illustration studying on graphs that makes use of graph convolutional neural networks and could be utilized constantly because the graph updates.

In addition to graph embeddings that present advanced vector representations, the brand new model of Neo4j for Graph Data Science provides basic machine studying algorithms such because the k-nearest neighbors algorithm (k-NN), generally used for pattern-based classification, to make it simpler to realize insights from graph embeddings.

Knowledge Graph Completion Example

Knowledge graph completion supplies worth throughout domains, together with figuring out new associations between genes and ailments, the invention of latest medicine and predicting hyperlinks between prospects and merchandise for higher suggestions. From queries to help area specialists in uncovering what they know, to patterns to grasp developments, to calculating high-value options to coach ML fashions, information graph completion is not potential with out graph know-how. 

In a drug discovery situation, this implies not solely figuring out potential new associations between genes and ailments or medicine and proteins but additionally offering quick context to evaluate the relevance or validity of those discoveries. For buyer suggestions, it means studying from consumer journeys to foretell correct suggestions for future purchases, whereas presenting choices inside their shopping for historical past to construct confidence in recommendations.

Find Out More

Learn more about Neo4j for Graph Data Science or obtain the latest version here.  
Version 1.four can be coated intimately throughout Neo4j Online Developer Expo and Summit (NODES) 2020, Neo4j’s world developer convention. You may watch the talks and demos from the Neo4j Connections for Graph Data Science on-line occasion.

About Neo4j

Neo4j is the chief in graph database know-how. As the world’s most generally deployed graph database, we assist world manufacturers – together with Comcast, NASA, UBS and Volvo – to disclose and predict how folks, processes and programs are interrelated. Using this relationships-first method, functions constructed with Neo4j deal with linked information challenges comparable to analytics and artificial intelligence, fraud detection, real-time recommendations and knowledge graphs. Find out extra at


Share this on Twitter

[email protected]

© 2020 Neo4j, Inc., Neo Technology®, Neo4j®, Cypher®, Neo4j® Bloom™, Neo4j® Aura™ and Neo4j for Graph Data Science™ are registered logos or a trademark of Neo4j, Inc. All different marks are owned by their respective firms.


Table of Contents

Related Links


Please enter your comment!
Please enter your name here