March 2022 Summaries

The International Consortium of Investigative Journalists (ICIJ) has released a massive investigation, dubbed the Pandora Papers, which reveals financial secrets of 35 current and former world leaders, over 330 politicians and public officials in 91 countries, and a worldwide roster of criminals. The investigation used Neo4j's graph databases to analyze the data they uncovered, allowing them to connect the dots between companies and individuals. This is ICIJ's latest investigation in a series that includes the Panama Papers and Paradise Papers, which exposed similar financial secrets and corruption across multiple countries. The Pandora Papers are notable for their complexity and global scope, involving 14 different firms with documents in multiple languages, and utilizing an open-source stack of technology to analyze the data.

Neo4j's AuraDB solution has been named "Cross Infrastructure Analytics Solution of the Year" in the third annual Data Breakthrough Awards program, marking the latest of 13 awards accumulated by Neo4j over the past year. This award recognizes the growing role Neo4j plays in data infrastructure for developers and data scientists. The company's marketing communications associate sat down with Kay Nair, Neo4j's VP of Global Cloud and Strategic Sales, to discuss AuraDB and its various versions, including Enterprise and Free, which has seen 500% year-over-year growth in enterprise customers. Notable customers include Levi Strauss & Co., NASA, and the UK Department of Education (DfE). The free tier of AuraDB allows users to create small projects without overhead, promoting innovation and community adoption. Neo4j plans to build out new security features and expand its cloud coverage in 2022, following this prestigious award as a validation of its early work and commitment to creating robust graph database-as-a-service (Graph DBaaS) solutions.

Building an information extraction pipeline allows developers to transform unstructured text inputs into useful knowledge graphs. This is achieved by applying Natural Language Processing (NLP) techniques, including coreference resolution, named entity recognition, and relationship extraction. The pipeline consists of four steps: inputting text, performing NLP techniques, storing the results in a graph database, and constructing a knowledge graph. The goal is to create a comprehensive knowledge graph that can be used for various applications, such as drug repurposing or analyzing biomedical concepts. By using NLP models trained on datasets like Wiki80, developers can extract relationships between entities and predict new use cases for existing drugs. This expands the scope of data that can be imported into a knowledge graph, making it more powerful and useful.

Neo4j GraphAcademy has released a new interactive course for Go developers called "How to build a Neo4j Application with Go", featuring hands-on training and covering the Neo4j Go Driver. This is part of GraphAcademy's free online training resource, which includes beginner courses and curated learning paths for various developer roles. The platform also offers a range of other resources, including training series, beginner blogs, under-the-hood explanations, and real-time dashboards. Users can participate in the Neo4j Developer Survey and catch up on past videos in the Bite-Sized Neo4j for Data Scientists series.

The presentation discusses the evolution of GraphQL and its integration with Neo4j, a graph database. The past version, neo4j-graphql.js, was an experiment to validate that GraphQL and Neo4j could work together, but it had limitations such as boilerplate code, developer productivity issues, extensibility problems, and performance concerns. The present version of the Neo4j GraphQL Library has been rewritten in TypeScript, providing better type safety, a comprehensive test suite, and compatibility with Neo4j's latest versions. It also introduces a new interface for relationship properties, allowing for more flexible and incremental growth of applications. Future features include cursor-based pagination, aggregations, unions and interfaces, indexes and constraints, and GraphQL subscriptions. The library is now stable and available for use, with an alpha release announced.

Neo4j is being used in various industries, including search for query expansion, recommendation engines, and visualizing hypergraphs of different subjects. Ashleigh Faith, a knowledge graph expert, has been using Neo4j for several years and has presented at the NODES Conference on its applications. She credits David Meza, Chief Knowledge Architect at NASA, with introducing her to Neo4j and inspiring her to pursue a PhD in the field. The Neo4j community is described as welcoming, with no such thing as a "stupid question." Ashleigh's favorite part of working with knowledge graphs is their ability to replicate human thought processes, making it easier for machines to understand complex concepts. She believes that graph technology will impact industries such as supply chain management, personal data protection, and water supply management in the coming years. Ashleigh has been honored with a Graphie Award, which she attributes to her passion for the field and her desire to demystify graph technology for others. She emphasizes the importance of showing the value of knowledge graphs rather than just presenting them as a solution.

This post presents a simple architecture for analyzing Bitcoin transactions in real-time using Neo4j and NeoDash. The authors utilize the Blockchain.com WebSocket API to listen for new transactions, which are then stored in a graph database (Neo4j) for analysis. A simplified graph model is used to capture the flow of value through the network, with addresses as nodes and transactions as relationships. The NeoDash tool is used to visualize data statistics and drill down into specific addresses or transactions. The project consists of three Docker services and can be easily set up using a `docker-compose` file. The authors demonstrate how to monitor overall transaction statistics, track address flows, and inspect individual transactions. The code is available on GitHub, and the authors invite readers to test it out and provide feedback, including any ideas for improving the code or exploring new use cases.

The article highlights the transformative influence of knowledge graphs in various fields, including medical research and crime investigations, by leveraging tools like Neo4j. The Viterbi algorithm is demonstrated as a powerful analytical tool for solving complex problems, while relationships between entities are visualized to gain insights. The article also showcases the use of Neo4j in language processing, bioinformatics, physics, finance, and other areas, emphasizing its ability to interactively inspect internal state values. Furthermore, it explores how knowledge graphs can be used to improve crime investigations through visualization and extract relationships from biomedical articles using NLP techniques. The article invites readers to participate in a developer survey, submit papers for GraphConnect 2022, and explore various tutorials and resources on Neo4j.

The Neo4j Quick Start template has been launched on Amazon Web Services (AWS) Marketplace, providing a streamlined deployment process for the Neo4j Graph Data Platform. This Quick Start allows users to deploy Neo4j in an automated and repeatable process, enabling a common architecture across organizations with a highly available distributed cluster that spans three Availability Zones. The template sets up managed network address translation gateways, Amazon EC2 Linux instances, Amazon Elastic Block Storage volumes, and preconfigures connectivity to the Cypher shell, Prometheus, Graphite, and JMX for day-two operation needs. This launch aims to empower Neo4j deployments with proven methods, enabling customers to streamline their deployment and operations of the Neo4j Graph Data Platform. The Quick Start also highlights the benefits of the Neo4j Graph Data Platform, including its native graph data store, data relationships, and graph exploration capabilities, as well as its support for data science and business applications.

Neo4j AuraDB Free is a free graph database that allows users to spin up their own database quickly and easily, with features such as the ability to visualize data in Neo4j Bloom and explore data in the Movie database. The platform offers a user-friendly experience, including Cypher commands for showing all data and deleting nodes and relationships. Users can start with the Movie database or bring in their own data, and get started with just a few minutes of setup time. With its free option and subscription to the Neo4j YouTube channel, users can access tons of graph tech goods and resources.

Neo4j, a company specializing in graph database technology, is highlighting the growing importance of knowledge graphs in businesses. Originally coined in 1972, knowledge graphs have gained significant attention over the past decade, particularly with Google's introduction of their use. The term refers to an interconnected dataset enriched with meaning, enabling complex decision-making and reasoning about underlying data. Knowledge graphs are being adopted by various organizations, including NASA, AstraZeneca, UBS, Lyft, Lockheed Martin, and Allianz, for applications such as skills discovery, patient journey analytics, product 360, and risk management. Recent webinars hosted by Neo4j have showcased the potential of knowledge graphs in supply chain management, digital twins, IoT, and other areas, offering benefits like end-to-end process views, agile problem resolution, and improved operational efficiency. Industry leaders and women are leading innovation with knowledge graphs, shaping the data landscape for years to come, as demonstrated by upcoming webinars on women-led innovation and applications of knowledge graphs in various industries.

The largest round of funding in database history was raised by Neo4j, a graph database company, with a valuation over $2 billion. This milestone marks the growth and acceptance of graph databases as a valuable category in the database market. The conversation highlights the explosive growth of graph databases, driven by their ability to handle complex connected data structures, which is becoming increasingly prevalent due to the world's growing connectivity. Neo4j's native graph architecture and its focus on developer adoption have been key factors in its success. The company has a large customer base, including Fortune 100 companies, and has shifted its go-to-market strategy from a bottom-up approach to targeting larger enterprises through cloud-based offerings. With its free tier and low-end pricing, Neo4j is expanding its addressable customer base, making it an attractive option for developers and data scientists worldwide.

Neo4j Under The Hood is a series of short videos presented by top engineering leaders, covering the power and transformative use cases of graph databases. Chris Gioran introduces fundamental concepts and showcases the nuts and bolts of graph technology. Graph databases are being used for fraud detection, where Zach Blumenfeld provides a 360-degree view of the workflow. Andrew Conacher demonstrates how native graph databases bring context to relationships in data, while Gonçalo Martins explains a scalable way to query data using Neo4j, Kafka, and the Outbox Pattern. Jennifer Reif shares her inspiration for engineering in music theory, and the community is encouraged to submit papers for GraphConnect 2022, participate in the Neo4j Developer Survey, and watch Bite-Sized Neo4j for Data Scientists series.

The text discusses the growing importance of knowledge graphs in leveraging data and delivering analytical insights, particularly with machine learning. Knowledge graphs are increasingly being adopted by organizations due to their flexibility, explicit storage of data relationships, and ability to surface context. The convergence of AI and knowledge graphs is driving intelligence into data, making it smarter and more trustworthy. This technology has significant advantages as graphs naturally store, compute, and analyze connections and relationships among data. Companies are leveraging knowledge graphs with machine learning for various use cases, from enhancing heuristics to complex uses like training embeddings in a graph-native learning model. The text highlights the benefits of combining knowledge graphs with AI, including improved accuracy, increased predictive power, and better decisioning capabilities.

Graph databases are designed to maintain arbitrary connections between data as a fully materialized entity that can be manipulated like any other piece of information in the database, making them dynamic and growing with the application. This allows for knowledge management and extraction, understanding, and utilization of data. Companies are adopting graph technologies to gain real-time analytics and decisioning capabilities. Graph databases excel at handling unstructured and changing data, making them suitable for various modern applications such as real-time recommendations, social network publishing, artificial intelligence, and more. Relational databases are not well-suited for the ever-changing needs of modern organizations due to their rigid data models and inability to adapt to change easily. The modern world generates data at exponential growth rates with dynamic connections, making relational systems brittle and unable to handle modern datasets. Graph databases offer a new foundation for data storage and manipulation with features like schema optional, index-free adjacency, and relationships as first-class citizens.

Neo4j has played a significant role in the innovative journey of OrbitMI, enabling them to create an artificial intelligence-based routing capability that provides fast and reliable results. The company's decision to build its own AI routing capability was driven by the need for scalability and performance, particularly when dealing with large datasets and complex use cases. With Neo4j, OrbitMI has been able to overcome challenges such as data complexity and spatial data support, allowing them to deliver a platform that provides value to its customers. The company believes that graph technology will continue to play a crucial role in supply chain management, enabling the efficient handling of multiple datasets and complex relationships between them. Overall, Neo4j's impact on OrbitMI has been transformative, providing the necessary tools and capabilities to drive innovation and growth.

The Neo4j GraphGists are teaching tools for developers that showcase "Use case and industry specific graph examples designed to inspire you towards your Graph epiphany." These GraphGists cover a wide range of industries, from pop culture and education, to sports and science. Anyone can submit a GraphGist to share with the Neo4j Community, which aims to highlight some of its favorites regularly, showcasing how graphs are used in various fields, such as deconstructing data from the International Ski Federation's alpine ski racing results using Cypher. The GraphGists portal provides inspiration for developers to explore graph technology, making it a valuable resource for those interested in learning more about Neo4j and its applications.

Graph embeddings are a new technology that learns the structure of connected data, revealing new ways to solve pressing problems and adding visibility to blind spots. They enable organizations to extract insight from their knowledge graphs, customer journeys, and transaction networks, providing predictive signals. Graph embeddings can distinguish normal behavior from anomalous transactions, identify duplicate users, improve product recommendations, discover new drugs, predict churn, and more. With the help of Neo4j Graph Data Science, businesses can put state-of-the-science techniques into production quickly, reliably, and at scale, empowering data scientists with less pain and extracting the full value from their data.

Neo4j is a graph database that stores relationships as first-class citizens, providing mission-critical context to data. It differs from traditional relational databases by storing data as graphs composed of nodes, labels, relationship types, and properties. In contrast, SQL-based databases require modeling relationships with additional tables and complex queries. Neo4j's approach enables effortless querying of relationships, reducing complexity and improving performance. The database allows users to specify relationship types in queries, providing a more natural way of expressing data connections.

Part 2 of the series on fraud detection using Neo4j and graph data science delves into identifying communities of users through entity resolution (ER) and the Weakly Connected Components (WCC) algorithm to enhance fraud detection. By establishing ER rules, such as linking users who share credit cards or devices with limited connections, the analysis creates relationships that help in identifying user communities. The WCC algorithm is then applied to these relationships to define communities, which are further analyzed to label accounts as fraud risks if they include flagged users. This process uncovered an additional 211 fraud risk accounts beyond the initially flagged 241, demonstrating the effectiveness of community-based analysis in fraud detection. The identified fraud risk accounts were involved in a significant portion of the peer-to-peer transaction amounts, indicating more extensive fraudulent networks than initially detected through chargeback analysis. The findings also suggest improvements in identifying fraudulent use of cards and devices, paving the way for further exploration of graph algorithms to identify more suspicious accounts and patterns.

In the realm of data science and machine learning, fraud detection remains a significant challenge due to the entities trying to prevent detection. However, graph-based approaches like Neo4j Graph Data Science can model relationships between entities, providing a powerful tool for rapidly exploring, analyzing, resolving, and predicting fraud entities and patterns. By applying these methods to an anonymized data sample from a peer-to-peer payment platform, it is possible to identify new fraud risks that went undetected with non-graph methods, increasing the number of flagged users by 87.5 percent. Furthermore, this approach can be highly scalable and transferable to various fraud detection use cases, enabling practitioners to build more accurate and sophisticated fraud detection applications.

This section explores how to apply graph machine learning to predict high fraud risk user accounts using Neo4j and Graph Data Science. The motivation behind this approach is to proactively detect fraudulent actors ahead of time, measure performance, automate the prediction of fraud risk accounts, and improve understanding of fraud patterns. The feature engineering strategy involves building features from previous analysis, including community indicators and size, PageRank on P2P with shared card degree, degree centrality on the shared id rule, other useful centrality features, and exporting these features to Python for training and evaluation of an ML model. A random forest classifier is used as a starting point, but other classifiers can be explored. The analysis demonstrates promising results with an accuracy of around 85% and provides insights into unlabeled high-probability fraud risk predictions, which require subject matter expert review and iteration to improve predictive performance.

The article explores the use of Neo4j and Graph Data Science (GDS) for fraud detection, specifically in identifying high-risk users and expanding on existing business logic to automatically identify suspiciously similar users. It leverages centrality and similarity algorithms to generate a ranked list of potential high-risk accounts using weighted degree centrality, which calculates the degree centrality of users based on their identifiers weighted by the fraudRiskRatio values. The article also uses node similarity algorithms to expand on existing communities of fraud risk users, identifying pairs of similar nodes based on Jaccard similarity calculations and projecting the graph to write relationships back to the database with a score representing similarity strength between user node pairs. These methods can help automate and semi-supervised processes for targeted triage and identification of suspicious user accounts based on previously labeled data.

In the first part of a series on fraud detection using Neo4j and graph data science, a sample graph dataset from a real-world peer-to-peer platform is introduced, featuring anonymized user accounts, transactions, and identifiers such as credit cards and devices. The dataset is structured to explore potential fraud patterns, with a small percentage of accounts flagged for fraud based on chargeback events and manual review. Initial analysis reveals that fraud accounts are not well-connected, suggesting possible limitations in the current labeling approach. To identify fragmented identities of fraudsters, community detection methods like Louvain are employed to partition the graph into well-connected groups. This exploratory analysis uncovers suspicious patterns, such as flagged users transferring money to non-flagged users with shared identifiers, which may indicate fraud risks. The article concludes with a preview of using these patterns to refine fraud detection in future installments of the series.