November 2018 Summaries

The U.S. retail industry is experiencing significant growth, with sales reaching trillions of dollars each year, and social media influencers playing a major role in shaping consumer behavior, generating billions of dollars. However, the market is also facing challenges due to constant tech disruptions and the increasing complexity of multi-channel commerce. To succeed, retailers must focus on mastering their connected data relationships, which can be leveraged through graph databases to better understand customer connections and preferences, ultimately delighting customers and gaining a competitive edge in this rapidly evolving retail landscape.

Neo4j 3.5 offers improved performance, utilization, and security for the Neo4j Graph Platform, a native graph database that is both DevOps- and DBA-friendly. The release features full-text indexing and search, faster sorting with index-backed ORDER BY, and a new Go language driver, among other capabilities. Native indexes have been completed, delivering dramatic performance improvements, while off-heap transaction state allows for better control of memory usage and more predictable behavior under heavy write loads. Security features include brute-force attack prevention, encryption of the cluster discovery service, and support for the X.509 standard extension for Subject Alternative Name (SAN) and Hostname verification.

At the time, he naturally knew that it would become a favorite metaphor used the world over to describe whenever we didn’t have a lot of something and then suddenly we had a lot of that same thing in great diversity. In fact, I’m sure this is precisely what he was thinking. The database ecosystem suddenly brimmed with diversity due to the Cambrian explosion of NoSQL technologies, making it impossible to cover the whole space in one blog post. This series will focus on the largest category within the NoSQL space: aggregate stores. Aggregate stores include key-value stores, wide-column stores, and document stores, which are distinguished from graph databases by their lack of connections between aggregates. Key-value stores store values organized by identifiers known as keys, with a basic model consisting of buckets that replicate onto several machines for fault-tolerance reasons. Wide-column stores have a sparsely populated table whose rows can contain arbitrary columns, providing natural indexing through keys. Document stores store and retrieve documents like an electronic filing cabinet, allowing for natural hierarchies and hierarchical JSON documents. While aggregate stores are good at storing big sets of discrete data, they sacrifice a data model, language, and functionality for handling data relationships, making them less suitable for problems that require understanding connectedness. Graph database technology, on the other hand, embraces relationships to solve problems involving context and connectedness.

The text discusses graph analytics, specifically traversal algorithms and graph properties that are essential to effectively innovate and develop intelligent solutions. It explains two fundamental graph traversal algorithms: breadth-first search (BFS) and depth-first search (DFS), highlighting their differences in exploring nodes in the graph. The article also covers various graph properties such as undirected vs. directed graphs, cyclic vs. acyclic graphs, weighted vs. unweighted graphs, and sparse vs. dense graphs. These properties are crucial in choosing the right traversal algorithm and ensuring efficient processing of the graph data. Additionally, the text touches on the importance of considering direction, cycles, weights, and density when analyzing graphs, and concludes by mentioning upcoming topics in the series, including Neo4j Graph Algorithms Library and pathfinding algorithms.

This week, we've picked a new video series to share with you from the Neo4j YouTube channel. The APOC library is a collection of user-defined procedures and functions created by the Neo4j community. The library contains over 1000 items, including gems that are often overlooked. To learn more about these hidden gems, check out the entire APOC video series. If you're interested in graph technology, be sure to subscribe to the Neo4j YouTube channel for new videos every week.

This week's Neo4j community highlights included online meetups on similarity algorithms and data ingestion using Kafka Streams and GraphQL. The Neo4j DevRel team attended events in Berlin and presented at the Data Natives conference. Featured community member Pat Patterson, Technical Director at StreamSets, was recognized for his contributions to the Neo4j community. Various blog posts were shared, including ones on building a session recommendation engine using Neo4j, GraphQL-based ingestion, and spatial support in neomodel. Additionally, a series of posts on decision streams and analysing PostgreSQL metadata using Neo4j were published. Upcoming events include iterative modeling of corporate resources in a rapidly growing company, while Neil Harbinger's tweet about getting ready for kerrydatascience with neo4j was shared as the tweet of the week.

StreamSets is using Neo4j to build flexible data pipelines that can handle continuous and streaming data from multiple sources, moving away from traditional ETL world of static schema. This allows companies to run their data flow operations in an efficient way with changing data and metadata. The choice of Neo4j was made due to its ability to integrate data from various sources and join it together for analyses across relationships. The future of graph technology is expected to see Neo4j becoming a data store in its own right, rather than just a point tool for visualization and analysis, making it an increasingly important part of StreamSets' work. Overall, Neo4j has been a pleasant experience to work with, offering stable and performant tools that enable the building of integrations with ease.

This presentation discusses the creation of a door-to-door journey-planning tool using Neo4j and Cypher, a query language for Neo4j. The author, Adam Cowley, wanted to create a tool that would allow him to accurately map a trip via train, bus, or both, and relied on open-source data from the UK government to compile both the locations and schedules of train and bus routes. He developed a 200-line Cypher query that pulled data from a database of more than 50,000 nodes with 75,000 relationships, with results returned within two to five seconds. The author discusses how he modeled the data using nodes and relationships, mined open data for train stations and schedules, and created a journey-planning tool that can handle both bus and train routes. He also shares his experience of implementing real-time data and overcoming issues such as deadlocks. The presentation concludes with the author's plan to use Java's Traversal API to put the journey-planning tool into production.

A graph platform, like Neo4j, is designed to facilitate collaboration among different groups in an organization by providing a variety of skill-specific tools for data scientists, solution developers, and business users. The platform offers efficient means for discovery and design, global analysis procedures, and algorithms that help identify communities, bottlenecks, influence points, and pathways. It also provides fast feedback loops, human-friendly ways to express connections, and a native graph database with analytics integration and robust procedures. Graph algorithms, such as those offered by Neo4j, reveal hidden patterns and structures in connected data, enabling pathfinding, centrality, and community detection, and are simple to apply across different user groups.

This week in Neo4j saw the presentation of Deep Learning on Graphs by Andrew Jefferson as part of an online meetup, highlighting the potential for neural networks to perform tasks on knowledge graphs. The team also announced major improvements to their Azure Cloud support, including new entries in the marketplace for Neo4j Enterprise and a revamped "High Availability" cluster entry. Additionally, Jennifer presented training classes and meetups on Graph Algorithms, while Alex Babeanu was featured as this week's community member for his work on using graphs for identity and access management. The team also released their official Neo4j Drivers, including Go, which brings the language in line with other languages in terms of features. Furthermore, Simonluca Landi wrote a blog post exploring the Lightning Network using graph algorithms, and the team is accepting proposals for talks at FOSDEM 2019.

Graph databases offer several advantages, including natural indexing by relationships, which enables users to include data without extensive modeling in advance. This makes graph technology particularly useful for fraud detection and analysis, where it can help identify patterns and anomalies that may indicate deception or speed. In this context, connected data analysis is especially beneficial, as it allows for the integration of behaviors and relationships to detect fraud. Graph technology can also be used to locate bot accounts in social media networks by depicting a graphical representation of the accounts with retweet counts as edge weights, which helps identify unnatural deviations and potential bots. Additionally, graph databases can be used to analyze circular payment patterns, where individuals send money to themselves to conceal it as legitimate transfers between normal accounts. Furthermore, graph technology can aid in detecting financial fraud and money laundering by applying classification, machine learning, pattern recognition, and statistical analysis to relationship models. By leveraging these capabilities, businesses can gain valuable insights from their data and stay ahead of scammers.

Neo4j is transitioning to an "open core" licensing model, where the Neo4j Enterprise Edition will be available solely under a commercial license, starting with the release candidate for version 3.5. This change aims to clarify the company's licensing model and provide a more industry-standard approach. The Neo4j Community Edition, however, will continue to be provided under an open source GPLv3 license, as it has been for years. The move does not affect customers, partners, OEMs, or Neo4j Desktop users. Some open source projects that rely on Neo4j Enterprise Edition may need to find alternative solutions, but the company is working with them to ensure a smooth transition. Despite this change, Neo4j remains committed to open source contributions and investments in various projects, including Neo4j Community Edition, openCypher, and others.

UBS bank's data distribution model was outdated, requiring manual joining of returned data to glean insights. The data services team created a platform that used metadata to automatically join data and provide pre-joined data to consumers. This was achieved using a graph database, which provided flexibility in modeling complex data flows and relationships between datasets. The knowledge base allowed users to understand what datasets were offered and how they related to each other, providing insights for both consumers and producers. The use of Neo4j as the graph database enabled easy querying and traversal of data, making it easier to answer questions about dataset attributes, transformations, and consumer views. Despite some challenges, including manual data loading and finding skilled personnel, the platform has been successful in delivering a self-service tool for browsing and requesting data distributions.

This blog series aims to help developers innovate and develop intelligent applications faster by leveraging graph analytics. Graph databases like Neo4j support numerous AI use cases, including knowledge graphs that represent knowledge in a form usable by machines. These graphs provide richer context for prescriptive analytics and AI applications, enabling intelligent applications like TextRank, NLP, and NLU technologies to get from text to meaning. Graphs are also used to feed machine learning models, find new features, and speed up AI decisions. Additionally, operational graphs can track AI decisions, providing a natural next step for intelligent applications. Graph analytics reveal global structures and patterns in data, enabling theory development through community detection and other algorithms. Human-friendly graph visualizations display or explain machine learning processes, accelerating data scientists' work and building confidence in AI solutions. Finally, graphs serve as a source of truth for all related AI components, creating a pipeline for iterative tasks and automating the sourcing and capture of related AI components.

The GraphCast series on the Neo4j YouTube channel offers short, educational videos about graph technology, including a music-inspired series called Graph Karaoke that features songs with graph database Cypher querying around song lyrics. The latest video in this series is "Heartbeat" by The Knife, which has an excellent Sunday groove and is available to watch on the Neo4j YouTube channel, along with other videos in the series featuring popular artists such as The Boss, Michael Jackson, AC/DC, and ABBA.

The Neo4j community has been actively sharing interesting content over the past seven days. Max De Marzi shared his talk on "Secret Sauce of Neo4j: Modeling and Querying Graphs," which explains how to get creative when modeling data with graph databases. Christophe Willemsen gave a preview of the full-text search capabilities that will be released in Neo4j 3.5, showcasing how to write Cypher queries using Lucene Classic Query Parser Syntax. The Neo4j DevRel team was busy sharing their love for graphs at conferences such as QCon and GraphQL Summit. Meanwhile, community members Jess Mason and Jason Cox were featured for running the Philly GraphDB meetup, which has grown to just short of 500 members. Other notable content includes a video on modeling a domain with Spring Data Neo4j and OGM, exploring systems landscape graph, node2vec, and new releases of Tableau Connector. Building graph apps with React is also covered in a blog post by Cristina Escalante, while Tom Zeppenfeldt created a video integrating the MeaningCloud topic extraction service with Graphileon. The next week's content is already looking promising, including a tweet from Karin Wolok about the buzzword status of Neo4j.

Comcast is using Neo4j to develop and launch new features rapidly, leveraging its flexibility in graph data modeling for persistent storage. The company adopted Neo4j four years ago after surveying the market and trying out various graph technologies. Its support model was a key factor in choosing Neo4j, as it provides a reliable back-end for production issues. Comcast has seen surprising results, including rapid innovation and delivery of features to customers, with applications such as personalized product recommendations and direct ticket purchases through the X1 Voice Remote. Looking ahead, Comcast believes that graph technology will become more prevalent in creating a data fabric, connecting siloed business units and unlocking new business opportunities.

The Graph DevRoom at FOSDEM 2019 is a one-day event focused on graph-related topics, including graph databases, query languages, and processing frameworks. The conference is part of the larger FOSDEM event, which brings together 8000+ free and open source enthusiasts from around the world. The Graph DevRoom will take place on February 2, 2019, with talks starting at 10:30am and a welcome at 10am. The event aims to bring together creators and maintainers of graph solutions, researchers, and open source hackers to share their latest work and learn from others. Proposals for talks or demos can be submitted by December 2, 2018, with important dates including submission deadlines and schedule publication. The conference is free to attend and promises to be a great experience.

Real-time recommendations are most effective when they consider both user needs and business strategies. This approach is being adopted by various industries beyond retail, travel, and entertainment, including government services, financial services, healthcare, and job recruiting. Graph technology, specifically the Neo4j Graph Platform, enables a hybrid recommender framework that leverages multiple techniques like collaborative filtering, content filtering, business rules, and knowledge-based filtering to provide accurate recommendations. This approach allows users to connect all their data and use multiple methods to gain control over the recommendation process.

The text discusses the development of Comcast's xFi smart home prototype, which aims to create a truly interactive and helpful home environment. The team identified key challenges, including semantic and social relationships between devices, people, and locations. To address this, they developed a rich definition of what a person and home are, using graph data structures and native graph databases. This allows for automation and personalization based on real-life relationships and context. The xFi prototype includes tools such as Kidwatch and Porchcam, which notify users when their children arrive home from school or when the doorbell rings. The system also uses GraphQL APIs to provide a flexible and expressive API for developers. Ultimately, the goal is to create a smart home that can learn and adapt to the user's needs, making it easier and more rewarding to live in a connected home environment.

This blog series explores how graph analytics can be used to innovate and develop intelligent solutions faster, with a focus on concrete use cases for graph technology. Graph technology is versatile and supports a broad range of use cases, including real-time fraud detection, personalized recommendations, and unified data views. It enables businesses to make timely operational decisions, streamline regulatory requirements, manage complex networks, and implement identity and access management systems. Additionally, it powers social applications and features by leveraging social connections and relationships in real-time.

This week in Neo4j saw significant funding, with the company closing an $80 million Series E round led by One Peak Partners and Morgan Stanley Expansion Capital. The funding will help Neo4j continue to deliver customer success with graph-powered business applications, including AI and machine learning systems. The company also released a new monitoring tool designed specifically for Neo4j administrators, called Halin. Additionally, the community featured Roland Guijt, a freelance software developer and trainer who updated his popular Pluralsight course on Graph databases and Neo4j to version 3.x. Other notable releases include gREST, a RESTful API development framework, and the neo4j-graphql.js library, which now includes new features for specifying type definitions. The community also discussed various use cases, including graph-enabled artificial intelligence and machine learning systems, as well as upcoming book releases, such as Graph-Powered Machine Learning by Dr. Alessandro Negro.

GraphAware is a company that specializes in helping organizations gain a better understanding of the knowledge spread across their organization. They use natural language processing to analyze and extract insights from unstructured data, such as text-based comments and stories, and integrate it with graph databases like Neo4j to create a collaborative knowledge graph. This allows companies to gather and organize data from multiple sources in a structured and homogenous way, creating new value for the organization by improving product quality, reducing costs, and delivering better services to end-users. GraphAware's technology, including their Hume platform, is expected to play a key role in the evolution of graph technology in the next five years, as more people move their data to graphs to extract insights and wisdom from it.

Neo4j has secured $80 million in a Series E funding round, raising its total growth funding to $160 million, the largest cumulative investment in the graph technology sector. The company's CEO, Emil Eifrem, attributes the need for this funding to the accelerating adoption of graph technology, which has moved from being a niche to a mainstream necessity for enterprises, especially for applications like fraud detection and recommendation engines. Neo4j plans to leverage this investment to enhance its graph platform, focusing on maintaining leadership in the space and expanding its offerings to meet increasing demand. A significant part of their strategy involves capitalizing on the synergy between graph technology and artificial intelligence, which they believe will lead to more accurate predictions and better decision-making as machine learning shifts from analyzing individual data points to connected data. This transition is expected to drive a new era of growth and innovation in graph-powered AI applications, presenting a substantial opportunity for Neo4j to seize.