Neo4j has released Graph Data Science Framework, a collection of Neo4j products designed to give data scientists tools to work with data containing highly predictive but underutilized relationships and network structures.
Neo4j database is one of the most popular graph databases. It stores data and relationships in graph structures, and is highly scalable. Developers can build intelligent applications that traverse s large, interconnected datasets in real time. It has a native graph storage and processing engine, and a graphical query language.
The new framework is made up of the Neo4j Graph Database, a graph science library, and Neo4j Bloom, a graph visualization and exploration tool that allows users to visualize algorithm results and find patterns without the need to write code.
The toolkit is described as having a flexible data structure for analytics and a library with five varieties of graph algorithms. Specifically, these start with community detection to detect group clustering or partition options.
Centrality or importance is used for assigning importance weights to distinct nodes in the network; while similarity evaluates how alike nodes are. Heuristic link prediction can be used to estimate the likelihood of nodes forming a relationship, while pathfinding & search provides the means to find optimal paths and to evaluate route availability and quality.
Neo4j Bloom is described as a graph visualization and exploration tool that allows users to visualize algorithm results and find patterns using codeless search. It has a ‘type to search’ interface where business users can type in typical values to identify categories, labels and relationships, and to inspect and edit data elements including properties, relationships and adjacent neighbors. The same interface can find and illustrate the shortest paths between nodes.
For more advanced users, it enables the definition of custom Cypher-based functions that can then be used by other less experienced users, and it’s possible to link from external.