December 2020 Summaries

NODES 2020 was a significant conference that brought together experts in graph databases and related fields. The event featured 73 great talks across six parallel tracks, covering various topics such as graph database internals, knowledge graphs, NLP, network science, and more. Notable speakers included Neo4j Chief Scientist Jim Webber, Lead Data Scientist at GraphAware Vlasta Kůs, and Google Cloud Developer Advocate Robert Kubis. The conference also showcased innovative applications of graph data science in domains like drug discovery, cyber security, and social good initiatives. With talks ranging from informative to fun, NODES 2020 provided a unique opportunity for attendees to learn about the latest advancements in graph databases and their potential impact on various industries.

Neo4j, a leading graph database provider, had a significant impact in 2020 with numerous releases and integrations that showcased its capabilities. Neo4j was recognized as a leader in The Forrester Wave for Graph Data Platforms, demonstrating its expertise in solving complex data and analytics problems. The company released Neo4j 4.0, which addressed scalability, intelligent data context, and robust enterprise-grade security, enabling enterprises to build intelligent applications that leverage interconnected data. Additionally, Neo4j was used by the International Consortium of Investigative Journalists to crack a vast industrial-scale money laundering case, highlighting its role in analyzing complex networks. The company also expanded its ecosystem with integrated graph database services on Google Cloud Platform and introduced connectors for Apache Spark and Business Intelligence tools. Furthermore, Neo4j announced Neo4j for Graph Data Science, the first enterprise graph framework for data scientists, which helps them leverage predictive power of relationships to answer unwieldy problems. Many leading companies, including NASA, ICIJ, ATPCO, Lyft, and others, chose Neo4j for their knowledge graphs, demonstrating its growing demand in industries such as supply chain management and business continuity.

Neo4j, the graph database company, had a remarkable year in 2020 with numerous exciting announcements, use cases, and community contributions. The company launched three new online training courses to help users learn about graph technology, including Neo4j 4.0. Graphs were also used to analyze the FinCEN Files, a global investigation that exposed industrial-scale money laundering, and to model Pokémon data. Knowledge graphs were discussed as essential for living in the big data era, with AI expert Charlie Beveridge describing their future significance. Machine learning algorithms and graph databases were explored in several blogs, including one on building an enterprise knowledge graph and another on financial fraud detection. The GraphConnect 2020 event was cancelled due to the COVID-19 pandemic, but its agenda is still available for those interested.

Neo4j, a leading graph database company, hosted various webinars in 2020 that covered topics such as human trafficking, responsible artificial intelligence, graph data science, and more. The webinars included a "Connections" series, which was a set of free monthly online conferences on graph technology; an introduction to Neo4j 4.0, the next-generation graph database built by developers for developers; a discussion on applying graphs to human trafficking crimes; a panel discussion on harnessing graphs in a new business climate; a session on improving insight into connected data using Linkurious Enterprise; an insider's guide to getting the most from graph data science in the enterprise; and a webinar on responsible AI practices, highlighting how graphs add context to guide more robust and trustworthy AI. These webinars showcased Neo4j's expertise in graph technology and its applications in various industries.

Neo4j is facing performance issues when making updates, particularly with large datasets. To optimize updates, Fanghua Joshua Yu, a Neo4j Pre-Sales & Field Engineer, reviews strategies such as using PERIODIC COMMIT with a small batchSize to reduce memory usage and parallel processing to improve performance. He also shares a case study of the Stack Overflow dataset, which has around 31 million nodes and 78 million relationships, to demonstrate how Cypher tuning can be used to keep queries efficient. Yu also discusses the importance of hardware, such as using an SSD, and monitoring system resources like heap memory usage and CPU threads. He highlights the benefits of using APOC procedures for iterative commits and parallel processing, which can significantly improve performance. Additionally, he emphasizes the need to test and profile updates before making large changes in a production environment.

To get started with graph data science, it's essential to investigate various use cases and become familiar with the concepts. Assembling a spearhead team of experts who can translate business needs into technical requirements is also crucial. Evaluating your "graphy" problem to identify areas where graph technology can solve interconnected information-dependent issues is vital. Next, assessing the current state of your organization or business and mapping the value of the proposed state are necessary steps. Measuring ROI, aligning stakeholders, getting project approval, conducting a proof of concept and planning for production, and connecting with the graph community will also help drive your graph project forward successfully. By following these tips and resources, you can successfully navigate the journey of implementing graph data science in your organization.

Graph databases are designed for highly interconnected datasets, mirroring the human brain's association model, and excel at identifying patterns and hidden connections. They offer a node-and-relationships format to efficiently store and query connected data, making them suitable for applications such as fraud detection, artificial intelligence, real-time recommendation engines, knowledge graphs, network and database infrastructure monitoring, and master data management. Graph databases overcome traditional relational database limitations by providing schema flexibility, adaptability, and the ability to tackle complex interdependencies, which are essential for modern business challenges. These technology choices have been adopted by numerous companies, including Fortune 500 entities, and are expected to further accelerate innovation in various industries.

The TWIN4j team released their last version for 2020, taking a break over Christmas and New Year's. They invite users to fill out the Developer Survey 2020, which will help them make better decisions in the future and earn $10 or donate it to charity. The team also published several videos from the NODES 2020 Extended series, including talks on manifold structures, Neo4j's role in recruiting, and building an election dashboard with low-code graph apps. Featured community member Klaus Blass is a consultant at the World Bank, working on CAPI surveys and censuses, and presented his work on converting census data to Neo4j graphs at NODES 2020. Other highlights include Will Lyon's blog post on building a podcast application with GRANDstack and Luanne Misquitta's insights on modeling IT Operations as a knowledge graph in Hume. Additionally, the team released new resources, including a sandbox for the New York Central Park data set and tutorials on running Neo4j on Windows Server 2019 Core Server.

Graph databases have gained significant attention in recent years due to their ability to handle large volumes of structured, unstructured, and semi-structured data. Unlike traditional relational databases, graph databases can efficiently explore complex relationships within data, making them ideal for applications such as fraud detection, recommendation systems, and network analysis. However, the adoption of graph databases is still relatively niche, with many companies viewing them as a specialized solution rather than a mainstream alternative to SQL databases. Despite this, researchers and developers are working to create an industry-wide standard query language for graph databases, which could lead to increased adoption and growth in the market. In the near future, it's unlikely that graph databases will replace SQL databases entirely, but they may coexist as complementary solutions, with graph databases being used for specific use cases where their unique strengths can provide significant benefits. Capgemini's UK Graph Guild is an initiative aimed at promoting the adoption of graph technology and providing a range of services and products to clients, including advice on choosing and implementing graph databases, developing graph-based proof-of-concepts, and offering graph-based investigative solutions such as Haystack.

Our names are Mark Needham and Amy Hodler, and we'll be discussing graph algorithms for community detection and recommendations. We'll explore Twitter's social graph, identify its influencers, and understand how to employ various graph algorithms. Specifically, we'll discuss Centrality algorithms like Degree, Closeness, Betweenness, and PageRank, as well as community detection algorithms like the Louvain Modularity algorithm. We'll also cover Cypher projections, which allow us to infer relationships in a graph and run different algorithms on them. Additionally, we'll introduce the Neo4j Graph Algorithms Library, which provides over 45 graph algorithms, including pathfinding, centrality, community detection, link prediction, and similarity algorithms. We'll also discuss how to use these algorithms for propagation pathways, flow and dynamics, and understanding group dynamics. Mark will talk about using graph algorithms in a more practical way, such as analyzing Twitter's social graph, identifying influencers, and running different algorithms on subsets of data. Amy will cover the basics of graph algorithms, including how they differ from querying, when to use them, and how to get started with Neo4j Graph Data Science Library. We'll also explore Cypher projections and NEuler, a React application that allows users to run graph algorithms without writing any Cypher code.

Financial institutions still struggle with costly manual efforts to patch gaps in their systems, resulting in higher operational expenditure (OPEX). Neo4j's graph technology was designed to resolve these gaps and complement existing solutions, particularly in the payments and transaction monitoring spaces. By automating complex calculations and entity resolution, Neo4j reduces OPEX spent on manual payment message repairs, detects suspicious transactions more effectively, and frees up resources for financial institutions to achieve a synchronized positive effect across all three operational spaces. With its graph database, Neo4j has proven effectiveness in reducing OPEX by up to 0.7% and improving the accuracy of transaction monitoring.

The Neo4j BI Connector is a tool that enables direct access to Neo4j graph data from business intelligence (BI) tools, allowing users to avoid coding and custom scripting. The connector works by transforming SQL queries into Cypher and fetching results, with an in-memory SQL Execution Engine that can handle standard SQL queries. Performance tuning and troubleshooting are crucial, as the driver may spill data to disk if it exceeds a certain memory limit, affecting query response times. Troubleshooting involves enabling query logging, using Neo4j monitoring tools, and analyzing logs for errors or excessive connections. The BI tooling layer also plays a role in troubleshooting, with log level adjustments providing insight into what the tool is doing. Overall, the Neo4j BI Connector enables powerful data querying and analysis capabilities in popular BI tools.

The Developer Relations team at Neo4j has published a new video titled "TWIN4j time" which covers various topics including GoGM, an object graph mapper for Neo4j and the Go programming language. The video features interviews with community members who have built projects using Neo4j such as Shyam Pratap Singh's Dark series graph database. Additionally, there are updates on community-driven projects like Lukas Böhres' use of Neo4j with PySpark and Tomaz Bratanic's network analysis of the ArXiv dataset to create a search and recommendation engine. The team also highlights featured community members such as Alex F. Mills who presented a talk at the Neo4j Connections event earlier this year.

The GraphCast series features a conversation between Rik Van Bruggen and Emil Eifrem, CEO and Co-Founder of Neo4j. They discuss the year's developments in graph technology, including shifting to the cloud, becoming more developer-centered, and the rise of graph data science. The conversation also touches on the potential of the property graph model and what the future holds for graph technology.

Graph data science (GDS) is increasingly applied in business to achieve better decisions, improved predictions, and innovation through the use of graph technology. Knowledge graphs are a fundamental component of GDS, providing a structured way to integrate diverse information and add context to applications such as AI systems. Graph analytics involves using global queries and algorithms to analyze relationships and topology within networks, enabling businesses to answer specific questions and make informed decisions. The major phases of the GDS journey include knowledge graph implementation, graph analytics, graph feature engineering, and graph embedding, which are used to extract predictive elements from raw graph data for machine learning tasks. Graph networks represent a new approach to ML that leverages graphs to improve results with less data, making predictions more explainable, and enabling new types of learning.

The GRANDstack is a full-stack framework for building applications with GraphQL, React, Apollo, and Neo4j Database. It makes handling complex data easy by modeling real-world data and relationships using Neo4j and GraphQL, allowing developers to create CRUD mutations automatically and override them as needed. The framework also enables the use of Cypher directives to add properties to GraphQL types and make the most of Neo4j. Additionally, GRANDstack allows for custom resolvers and queries in APOC and Cypher to create what's needed, making it easier to model real-world data such as oil and gas ownership relationships with many different types of ownership and directions of relationships involved.

The Neo4j Graphies recognize excellence in graph technology across various use cases and industries, showcasing cutting-edge or impactful graph-based projects. The awards celebrate innovative individuals and organizations using Neo4j graph databases to make positive change in the world. Winners include several Neo4j customers, such as Allianz Benelux, Danish Business Authority, and Meredith Corporation, who have implemented graph technology for customer profiling, fraud detection, machine learning, and more. The Graphies also recognize individual contributors, including Elena Kohlwey and Reshama Shaikh, who are part of the Neo4j community and have made significant contributions to the field. The awards aim to promote awareness of the impact of graph technology and provide a platform for thought leaders to showcase their work.

The Neo4j BI Connector provides direct access to Neo4j graph data from business intelligence tools, delivering connected data results without requiring coding or custom scripting. The connector offers a secure way to manage access to graph data by segregating it into multiple graphs and controlling access within each graph through fine-grained access control. This security configuration applies to the BI Connector itself, ensuring that users can only see the schema of the Neo4j graph but not the actual data unless granted permission. By centralizing access management using existing Neo4j techniques, external users can be given their own user accounts and passwords, while graph administrators can grant privileges to only necessary databases, resulting in a more secure and controlled environment for managing access to graph data.

Gerrit Meier launched a new series about Spring Data Neo4j, while Adam Cowley announced the launch of Charts, a Neo4j Graph App that makes it easy to create charts on top of graph data. Hantsy continued the Spring Data Neo4j series by showing how to write custom queries using the @Query annotation in Spring Data Neo4j. Meanwhile, Jesús Barrasa built a graph of the UNESCO Thesaurus and Elaine Rosenberg launched the new Neo4j 4.0 Certification exam. The Developer Relations team also shared news about various tools and updates, including BruteShark's version v1.1.2 and Linkurious Enterprise's version 2.9.10. Additionally, Chuck Frisbie was featured as a community member, having recently presented Turbocharging Advanced Analytics with Sisense and the Neo4j Connector for BI at the Neo4j Connections event.

I was working on a project trying to solve a specific hierarchical problem and found GraphX, other solutions, and then Neo4j. I researched it and downloaded a book that gave a good introduction to the product and its founder's story. After 10 years in data visualization engineering, I realized most problems were graphy problems, which excites me as it opens up technology to more people beyond developers. I wanted to meet others using Neo4j, see use cases, and understand next steps, making GraphTour a good opportunity for me.

Neo4j, a leading Graph Data Platform as recognized by Forrester Research, has undergone significant development since its inception, learning valuable lessons along the way. The company's pragmatic property graph model was first sketched on an airplane napkin and later implemented successfully in Sweden. Neo4j's focus has always been on making developers' lives easier, which led to the creation of tools like Cypher query language and Neo4j Browser. The platform has evolved significantly, handling transactions, memory, and I/O like a proper database. Its scalability is top-rated, with performance critical for transactional use cases, and it can scale from single instances to clustered environments. Despite its global presence, Neo4j prioritizes data safety and security, ensuring the trust of its users and customers. The company welcomes competition from giants like Amazon and Microsoft, seeing their entry as a validation of the graph market segment and growth for the visibility of graph technology. Future plans include democratizing graph data science, making Neo4j available on all cloud platforms, and continuing to make it easier to build applications and import data into cloud-native services.

The GraphAcademy site has introduced a new certification exam for Neo4j, focusing on features in release 4.0, including Cypher changes and improvements, multi-database management, role-based access control, and fabric. The exam consists of 30 questions that must be answered within 45 minutes, with a minimum score of 80% required to pass. To prepare for the exam, users can utilize various resources such as online courses and Neo4j manuals, which cover topics like Cypher subqueries, administration, and fabric configuration. The certification is free and can be taken multiple times within a 24-hour period until passing is achieved.

This blog post uses Neo4j and GRANDstack to build a graph and timeline visualization app that drives fraud investigations. It focuses on fraud in Wikipedia articles about the 2020 U.S. election presidential candidates, exploring why an anomaly detection platform featuring dedicated timeline analysis is an effective choice. The author loads Wikipedia data into Neo4j using the MediaWiki API, then visualizes it with ReGraph and KronoGraph to detect potentially fraudulent behaviors. The app features a network view of connections between editors and pages, as well as two fully-interactive time-based components: the ReGraph time bar at the bottom of the screen and KronoGraph timelines to the right of the network. These tools allow for the exploration of editing patterns, recognition of outliers, and spotting periods of intense activity, ultimately driving fraud investigations with smarter analytics.