Neo4j, profile picture
Uploaded byNeo4j
PDF, PPTX1,801 views

Improving Machine Learning using Graph Algorithms

The document discusses enhancing machine learning (ML) predictions through graph algorithms, emphasizing the importance of relationships in predicting behavior. It highlights features extraction and engineering from graph databases like Neo4j, which provide novel insights and improve ML models by utilizing connected data structures. Various use cases, such as predicting collaboration in academic networks and applications in industries like healthcare and finance, illustrate the advantages of integrating graph algorithms into ML workflows.

Embed presentation

Download as PDF, PPTX
Improve ML Predictions using Graph Algorithms Mark Needham, Neo4j Amy Hodler, Neo4j May 2019 #Neo4j #GraphAnalytics
• Graphs for Predictions • Connected Features • Link Prediction • Neo4j + Spark Workflow Amy E. Hodler Graph Analytics & AI Program Manager, Neo4j Amy.Hodler@neo4j.com @amyhodler neo4j.com/ graph-algorithms-book Chapter 8: Graph + ML Spark & Neo4j Mark Needham Developer Relations Engineer, Neo4j Mark.needham@neo4j.com @markHneedham
What in Common is Predictive?
Relationships: Strongest Predictors of Behavior! “Increasingly we're learning that you can make better predictions about people by getting all the information from their friends and their friends’ friends than you can from the information you have about the person themselves” James Fowler David Burkus James Fowler Albert-Laszlo Barabasi
Native Graph Platforms are Designed for Connected Data TRADITIONAL PLATFORMS BIG DATA TECHNOLOGY Store and retrieve data Aggregate and filter data Connections in data Real time storage & retrieval Real-Time Connected Insights Long running queries aggregation & filtering “Our Neo4j solution is literally thousands of times faster than the prior MySQL solution, with queries that require 10-100 times less code” Volker Pacher, Senior Developer Max # of hops ~3 Millions
Graph Database Surging in Popularity Trends since Jan 2013 DB-Engines.com
Graph Data Science Applications
• Current data science models ignore network structure & complex relationships • Graphs add highly predictive features to existing ML models • Otherwise unattainable predictions based on relationships Novel & More Accurate Predictions with the Data You Already Have Machine Learning Pipeline
Connected Features
Connection-related metrics about our graph, such as the number of relationships going into or out of nodes, a count of potential triangles, or neighbors in common. 14c What Are Connected Features?
Query (e.g. Cypher) Real-time, local decisioning and pattern matching Graph Algorithms Libraries Global analysis and iterations You know what you’re looking for and making a decision You’re learning the overall structure of a network, updating data, and predicting Local Patterns Global Computation Deriving Connected Features
Connected Feature Engineering Feature Engineering is how we combine and process the data to create new, more meaningful features, such as clustering or connectivity metrics. Add More Descriptive Features: - Influence - Relationships - Communities Extraction
17 Graph Feature Categories & Algorithms Pathfinding & Search Finds the optimal paths or evaluates route availability and quality Centrality / Importance Determines the importance of distinct nodes in the network Community Detection Detects group clustering or partition options Heuristic Link Prediction Estimates the likelihood of nodes forming a relationship Evaluates how alike nodes are Similarity Embeddings Learned representations of connectivity or topology
Link Prediction
19 Can we infer new interactions in the future? What unobserved facts we’re missing?
+ 50 years of biomedical data integrated in a knowledge graph Predicting new uses for drugs by using the graph structure to create features for link prediction Example: het.io
Example: het.io
Methods for Link Prediction Algorithm Measures Run targeted algorithms and score outcomes Set a threshold value used to predict a link between nodes Machine Learning Use the measures as features to train an ML model Community Detection Link Prediction Similarity 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 2 4 15 1 3 4 7 12 1 5 6 1 1 0 Centrality
Example: Predicting Collaboration
• Citation Network Dataset - Research Dataset – “ArnetMiner: Extraction and Mining of Academic Social Networks”, by J. Tang et al – Used a subset with 52K papers, 80K authors, 140K author relationships and 29K citation relationships • Neo4j – Create a co-authorship graph and connected feature engineering • Spark and MLlib – Train and test our model using a random forest classifier 24 Predicting Collaboration with a Graph Enhanced ML Model
Our Link Prediction Workflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize
Our Link Prediction Workflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identified sparse feature areas Feature Engineering: New graphy features
Graph Algorithms Used for Feature Engineering (few examples) Preferential Attachment measure the closeness of nodes based on shared neighbors Common Neighbors measures the number of possible neighbors (triadic closure) Illustration be.amazd.com/link-prediction/
Graph Algorithms Used for Feature Engineering (few examples) Triangle counting and clustering coefficients measure the density of connections around nodes Louvain Modularity identifies interacting communities and hierarchies
Our Link Prediction Workflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identified sparse feature areas Feature Engineering: New graphy features Train / Test Split Resample: Downsampled for proportional representation
31
32 Test/Train Split 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 2 4 15 1 3 4 7 12 1 5 6 1 1 0 2 12 3 3 0 4 9 4 8 1 7 10 12 36 1 8 11 2 3 0
33 Test/Train Split 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 2 4 15 1 3 4 7 12 1 5 6 1 1 0 2 12 3 3 0 4 9 4 8 1 7 10 12 36 1 8 11 2 3 0 Train Test
OMG I’m Good! Data Leakage! Graph metric computation for the train set touches data from the test set. Did you get really high accuracy on your first run without tuning?
Train and Test Graphs: Time Based Split 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 2 4 15 1 3 4 7 12 1 5 6 1 1 0 Train Test 1st Node 2nd Node Common Neighbors Preferential Attachment label 2 12 3 3 0 4 9 4 8 1 7 10 12 36 1 < 2006 >= 2006
Train and Test Graphs: Time Based Split 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 2 4 15 1 3 4 7 12 1 5 6 1 1 0 Train Test 1st Node 2nd Node Common Neighbors Preferential Attachment label 2 12 3 3 0 4 9 4 8 1 7 10 12 36 1
Class Imbalance Negative Examples Positive Examples
There are significantly more negative examples than positive ones: # negative examples = (# nodes)² - (# relationships) - (# nodes) 38 Class Imbalance
A very high accuracy model could predict that a pair of nodes are not linked. 39 Class Imbalance
Class Imbalance
Our Link Prediction Workflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identified sparse feature areas Feature Engineering: New graphy features Train / Test Split Resample: Downsampled for proportional representation Model Selection: Random Forest Ensemble method
Picking a Classifier
Training Our Model This is one decision tree in our Random Forest used as a binary classifier to learn how to classify a pair: predicting either linked or not linked.
4 Models Trained with Multiple Graph Features Graph Features: • Common Authors “Graphy” Model Common Authors Model Triangles Model Community Model Graph Features: • Preferential Attachment • Total Neighbors Graph Features: • Min & Max Triangles • Min & Max Clustering Coefficient Graph Features: • Label Propagation • Louvain Modularity
Our Link Prediction Workflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identified sparse feature areas Feature Engineering: New graphy features Train / Test Split Resample: Downsampled for proportional representation Precision, Accuracy, Recall ROC Curve & AUC Model Selection: Random Forest Ensemble method
Measures Accuracy Proportion of total correct predictions. Beware of skewed data! Precision Proportion of positive predictions that are correct. Low score = more false positives Recall / True Positive Rate Proportion of actual positives that are correct. Low score = more false negatives False Positive Rate Proportion of incorrect positives ROC Curve & AUC X-Y Chart mapping above 2 metrics (TPR and FPR) with area under curve
Result: First Model ROC & AUC Problematic False Positives! Common Authors Model 1
Result: All Models Common Authors Model 1 Community Model 4
Iteration & Tuning: Feature Influence For feature importance, the Spark random forest averages the reduction in impurity across all trees in the forest Feature rankings are in comparison to the group of features evaluated Also try PageRank! Try removing different features (LabelPropagation)
Graph Machine Learning Workflow Data aggregation Create and store graphs Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identify uninteresting features Cleanse (outliers+) Feature engineering/ extraction Train / Test split Resample for meaningful representation (proportional, etc.) Precision, accuracy, recall (ROC curve & AUC) SME Review Cross-validation Model & variable selection Hyperparameter tuning Ensemble methods
Resources • neo4j.com/sandbox • neo4j.com/developer/ graph-algorithms/ • community.neo4j.com Data & Code: • This example from O’Reilly book bit.ly/2FPgGVV (ML Folder) Amy.Hodler@neo4j.com @amyhodler neo4j.com/ graph-algorithms-book
Q&A/Extra Stuff to delete 52
53 Connected Feature Extraction Feature Extraction is how when we change the shape or format of the data to be usable in a machine learning pipeline. For example, from a graph, we extract the relevant subset of the data into a tabular format for model building.
Connected Feature Selection Feature Selection is how we reduce the number of features used in a model to a relevant subset. This can be done algorithmically or based on domain expertise, but the objective is to maximize the predictive power of your model while minimizing overfitting.
720+ 7/10 12/2 5 8/10 53K+ 100+ 300+ 450+ Adoption Top Retail Firms Top Financial Firms Top Software Vendors Customers Partners • Creator of the Neo4j Graph Platform • ~250 employees • HQ in Silicon Valley, other offices include London, Munich, Paris and Malmö Sweden • $80M new funding led by Morgan Stanley & One Peak. Total $160M from Fidelity, Sunstone, Conor, Creandum, and Greenbridge Capital • Over 15M+ downloads & container pulls • 325+ enterprise subscription customers with over half with >$1B in revenue Ecosystem Startups in program Enterprise customers Partners Meet up members Events per year Industry’s Largest Dedicated Investment in Graphs Neo4j - The Graph Company
Strictly ConfidentialStrictly Confidential 56 Helping The World To Make Sense of Data ICIJ used Neo4j to uncover the world’s largest journalistic leak to date, The Panama Papers NASA uses Neo4j for a “Lessons Learned” database to improve effectiveness in search missions in space Neo4j is used to graph the human body, map correlations, identify cause & effect and search for the cure for cancer SAVING DEMOCRACY MISSION TO MARS CURING CANCER
Graph and ML Algorithms in Neo4j • Parallel Breadth First Search & DFS • Shortest Path • Single-Source Shortest Path • All Pairs Shortest Path • Minimum Spanning Tree • A* Shortest Path • Yen’s K Shortest Path • K-Spanning Tree (MST) • Random Walk • Degree Centrality • Closeness Centrality • CC Variations: Harmonic, Dangalchev, Wasserman & Faust • Betweenness Centrality • Approximate Betweenness Centrality • PageRank • Personalized PageRank • ArticleRank • Eigenvector Centrality • Triangle Count • Clustering Coefficients • Connected Components (Union Find) • Strongly Connected Components • Label Propagation • Louvain Modularity – 1 Step & Multi-Step • Balanced Triad (identification) • Euclidean Distance • Cosine Similarity • Jaccard Similarity • Overlap Similarity • Pearson Similarity Pathfinding & Search Centrality / Importance Community Detection Similarity neo4j.com/docs/ graph-algorithms/current/ Updated April 2019 Link Prediction • Adamic Adar • Common Neighbors • Preferential Attachment • Resource Allocations • Same Community • Total Neighbors
Conceive Code Compute Store Non-Native Graph DBNative Graph DB RDBM S Optimized for graph workloads Connectedness Differentiates Neo4j
Neo4j is an enterprise-grade native graph platform that enables you to: • Store, reveal and query data relationships • Traverse and analyze any levels of depth in real-time • Add context and connect new data on the fly 59 Who We Are: Leader in Graph Innovations • Performance • ACID Transactions • Schema-free Agility • Graph Algorithms Designed, built and tested natively for graphs from the start for: • Developer Productivity • Hardware Efficiency • Global Scale • Graph Adoption Graph Transactions Graph Analytics Data Integration Development & Admin Analytics Tooling Drivers & APIs Discovery & Visualization
60 • Record “Cyber Monday” sales • About 35M daily transactions • Each transaction is 3-22 hops • Queries executed in 4ms or less • Replaced IBM Websphere commerce • 300M pricing operations per day • 10x transaction throughput on half the hardware compared to Oracle • Replaced Oracle database • Large postal service with over 500k employees • Neo4j routes 7M+ packages daily at peak, with peaks of 5,000+ routing operations per second. Handling Large Graph Work Loads for Enterprises Real-time promotion recommendations Marriott’s Real-time Pricing Engine Handling Package Routing in Real-Time
Recommendations Dynamic Pricing IoT-applicationsFraud Detection Real-Time Transaction Applications Generate and Protect Revenue Customer Engagement Metadata and Advanced Analytics Data Lake Integration Knowledge Graphs for AI Risk Mitigation Generate Actionable Insights Network Management Supply Chain Efficiency Identity and Access Management Internal Business Processes Improve Efficiency and Cut Costs Graph Use Cases by Value Proposition
Softwar e Financial Services Teleco m Retail & Consumer Goods Media & Entertainment Other Industries Airbus 62 Copyright © 2017 Neo4j, Inc. Company Confidential
Graph Transactions Graph Analytics Data Integration Development & Admin Analytics Tooling Drivers & APIs Discovery & Visualization Developers Admins Applications Business Users Data Analysts Data Scientists Enterprise Data Hub Native Graph Platform: Tools for Many Users
Collections-Focused Multi-Model, Documents, Columns & Simple Tables, Joins Neo4j is designed for data relationships Different Paradigms NoSQL Relational DBMS Neo4j Graph Platform Connections-Focused Focused on Data Relationships Development Benefits Easy model maintenance Easy query Deployment Benefits Ultra high performance Minimal resource usage
How Neo4j Fits — Common Architecture Patterns From Disparate Silos To Cross-Silo Connections From Tabular Data To Connected Data From Data Lake Analytics to Real-Time Operations
Cypher: Powerful & Expressive Query Language MATCH (:Person { name:“Dan”} ) -[:MARRIED_TO]-> (spouse) MARRIED_TO Dan Ann NODE RELATIONSHIP TYPE LABEL PROPERTY VARIABLE
Neo4j Bloom 67 • High fidelity • Scene navigation • Property views • Search suggestions • Saved phrase history • Property editor • Schema perspectives • Bloom chart type • Visualize • Communicate • Discover • Navigate • Isolate • Edit • Share
68 Real-Time Recommendations Fraud Detection Network & IT Operations Master Data Management Knowledge Graph Identity & Access Management Common Graph Technology Use Cases AirBnb
Graphs Drive Innovation 69 Context Paths Auto-Graphs Graph Layers 1st Graph Cross- Connect Cross-tech applications Internet of Things operations Transparent Neural Networks Blockchain-managed systems Adjacent graph layers inspire new innovations Metadata / Risk Management Knowledge Graphs AI- Powered Customer Experiences Connect unlike objects such as people to products, locations Mobile app explosion Recommendation engines Fraud detectors Desire for more context to follow connections Connects like objects People, computer networks, telco, etc
Business Problem • Find relationships between people, accounts, shell companies and offshore accounts • Journalists are non-technical • Biggest “Snowden-Style” document leak ever; 11.5 million documents, 2.6TB of data Solution and Benefits • Pulitzer Prize winning investigation resulted in robust coverage of fraud and corruption • PM of Iceland & Pakistan resigned, exposed Putin, Prime Ministers, gangsters, celebrities (Messi) • Led to assassination of journalist in Malta Background • International Consortium of Investigative Journalists (ICIJ), small team of data journalists • International investigative team specializing in cross-border crime, corruption and accountability of power • Works regularly with leaks and large datasets ICIJ Panama Papers INVESTIGATIVE JOURNALISM Fraud Detection / Knowledge Graph70
Thomson Reuters Graph 71 • Data Fusion for Portfolio Managers • Graph layers
Background • Personal shopping assistant • Converses with buyer via text, picture and voice to provide real-time recommendations • Combines AI and natural language understanding (NLU) in Neo4j Knowledge Graph • First of many apps in eBay's AI Platform Business Problem • Improve personal context in online shopping • Transform buyer-provided context into ideal purchase recommendations over social platforms • "Feels like talking to a friend" Solution and Benefits • 3 developers, 8M nodes, 20M relationships • Needed high-performance traversals to respond to live customer requests • Easy to train new algorithms and grow model • Generating revenue since launch eBay for Google Assistant ONLINE RETAIL Knowledge Graph powers Real-Time Recommendations72 EE Customer since 2016 Q3
Background • Over 7M citizens suffer from Diabetes • Connecting over 400 researchers • Incorporates over 50 databases, 100k’s of Excel workbooks, 30 database of biological samples • Sought to examine disease from as many angles as possible. Business Problem • Genes are connected by proteins or to metabolites, and patients are connected with their diets, etc… • Needed to improve the utilization of immensely technical data • Needed to cater to doctors and researchers with simple navigation, communication and connections of the graph. Solution and Benefits • Dr. Alexander Jarasch, Head of Bioinformatics and Data Management • Scientists can conduct parallel research without asking the same questions or repeating tests • Built views like a liver sample knowledge graph DZD - German Center for Diabetes Research Medical Genomic Research73 EE Customer since 2016 Q4

More Related Content

PDF
Graph-Powered Machine Learning
byDatabricks
 
PDF
Workshop - Neo4j Graph Data Science
byNeo4j
 
PDF
GPT and Graph Data Science to power your Knowledge Graph
byNeo4j
 
PDF
How Powerful are Graph Networks?
byIAMAl
 
PDF
Slides: Knowledge Graphs vs. Property Graphs
byDATAVERSITY
 
PPTX
Boost Your Neo4j with User-Defined Procedures
byNeo4j
 
PPTX
Introduction to Graph Databases
byMax De Marzi
 
PDF
Introduction to Graph Databases
byDataStax
 
Graph-Powered Machine Learning
byDatabricks
 
Workshop - Neo4j Graph Data Science
byNeo4j
 
GPT and Graph Data Science to power your Knowledge Graph
byNeo4j
 
How Powerful are Graph Networks?
byIAMAl
 
Slides: Knowledge Graphs vs. Property Graphs
byDATAVERSITY
 
Boost Your Neo4j with User-Defined Procedures
byNeo4j
 
Introduction to Graph Databases
byMax De Marzi
 
Introduction to Graph Databases
byDataStax
 

What's hot

PDF
3. Relationships Matter: Using Connected Data for Better Machine Learning
byNeo4j
 
PDF
Relationships Matter: Using Connected Data for Better Machine Learning
byNeo4j
 
PDF
Knowledge Graphs and Generative AI
byNeo4j
 
PDF
Graph based data models
byMoumie Soulemane
 
PDF
Optimizing the Supply Chain with Knowledge Graphs, IoT and Digital Twins_Moor...
byNeo4j
 
PDF
Neo4j Graph Data Science - Webinar
byNeo4j
 
PDF
Graph database Use Cases
byMax De Marzi
 
PDF
Deep Learning for Graphs
byDeepLearningBlr
 
PDF
Natural Language Processing with Graph Databases and Neo4j
byWilliam Lyon
 
PDF
Introduction of Knowledge Graphs
byJeff Z. Pan
 
PDF
Neo4j in Depth
byMax De Marzi
 
PDF
Data Modeling with Neo4j
byNeo4j
 
PPTX
Neo4j GraphSummit London March 2023 Emil Eifrem Keynote.pptx
byNeo4j
 
PPTX
Introduction to Graph neural networks @ Vienna Deep Learning meetup
byLiad Magen
 
PDF
Neo4j Demo: Using Knowledge Graphs to Classify Diabetes Patients (GlaxoSmithK...
byNeo4j
 
PDF
Introducing Neo4j
byNeo4j
 
PDF
Webinar on Graph Neural Networks
byLucaCrociani1
 
PPTX
Graph databases
byVinoth Kannan
 
PPTX
Graphs in Automotive and Manufacturing - Unlock New Value from Your Data
byNeo4j
 
PPTX
Intro to Neo4j
byNeo4j
 
3. Relationships Matter: Using Connected Data for Better Machine Learning
byNeo4j
 
Relationships Matter: Using Connected Data for Better Machine Learning
byNeo4j
 
Knowledge Graphs and Generative AI
byNeo4j
 
Graph based data models
byMoumie Soulemane
 
Optimizing the Supply Chain with Knowledge Graphs, IoT and Digital Twins_Moor...
byNeo4j
 
Neo4j Graph Data Science - Webinar
byNeo4j
 
Graph database Use Cases
byMax De Marzi
 
Deep Learning for Graphs
byDeepLearningBlr
 
Natural Language Processing with Graph Databases and Neo4j
byWilliam Lyon
 
Introduction of Knowledge Graphs
byJeff Z. Pan
 
Neo4j in Depth
byMax De Marzi
 
Data Modeling with Neo4j
byNeo4j
 
Neo4j GraphSummit London March 2023 Emil Eifrem Keynote.pptx
byNeo4j
 
Introduction to Graph neural networks @ Vienna Deep Learning meetup
byLiad Magen
 
Neo4j Demo: Using Knowledge Graphs to Classify Diabetes Patients (GlaxoSmithK...
byNeo4j
 
Introducing Neo4j
byNeo4j
 
Webinar on Graph Neural Networks
byLucaCrociani1
 
Graph databases
byVinoth Kannan
 
Graphs in Automotive and Manufacturing - Unlock New Value from Your Data
byNeo4j
 
Intro to Neo4j
byNeo4j
 

Similar to Improving Machine Learning using Graph Algorithms

PDF
Transforming AI with Graphs: Real World Examples using Spark and Neo4j
byDatabricks
 
PPTX
Graph Analytics: Graph Algorithms Inside Neo4j
byNeo4j
 
PDF
Knowledge graphs, meet Deep Learning
byConnected Data World
 
PPTX
How Graphs are Changing AI
byNeo4j
 
PDF
Graph Algorithms for Developers
byNeo4j
 
PDF
Predicting Influence and Communities Using Graph Algorithms
byDatabricks
 
PPTX
Graphs for Ai and ML
byNeo4j
 
PDF
Scalable Graph Convolutional Network Based Link Prediction on a Distributed G...
bymiyurud
 
PDF
Improve ML Predictions using Connected Feature Extraction
byDatabricks
 
PDF
Improve ml predictions using graph algorithms (webinar july 23_19).pptx
byNeo4j
 
PDF
Improve ML Predictions using Graph Analytics (today!)
byNeo4j
 
PPTX
GraphTour Boston - Graphs for AI and ML
byNeo4j
 
PDF
Leveraging Graphs for Better AI
byNeo4j
 
PDF
GraphTour London 2020 - Graphs for AI, Amy Hodler
byNeo4j
 
PDF
IRJET- Link Prediction in Social Networks
byIRJET Journal
 
PDF
Leveraging Graphs for Better AI
byNeo4j
 
PDF
GraphTour 2020 - Graphs & AI: A Path for Data Science
byNeo4j
 
PDF
How Graphs Enhance AI
byNeo4j
 
PDF
Transforming AI with Graphs: Real World Examples using Spark and Neo4j
byFred Madrid
 
PDF
Spark Summit EU talk by Reza Karimi
bySpark Summit
 
Transforming AI with Graphs: Real World Examples using Spark and Neo4j
byDatabricks
 
Graph Analytics: Graph Algorithms Inside Neo4j
byNeo4j
 
Knowledge graphs, meet Deep Learning
byConnected Data World
 
How Graphs are Changing AI
byNeo4j
 
Graph Algorithms for Developers
byNeo4j
 
Predicting Influence and Communities Using Graph Algorithms
byDatabricks
 
Graphs for Ai and ML
byNeo4j
 
Scalable Graph Convolutional Network Based Link Prediction on a Distributed G...
bymiyurud
 
Improve ML Predictions using Connected Feature Extraction
byDatabricks
 
Improve ml predictions using graph algorithms (webinar july 23_19).pptx
byNeo4j
 
Improve ML Predictions using Graph Analytics (today!)
byNeo4j
 
GraphTour Boston - Graphs for AI and ML
byNeo4j
 
Leveraging Graphs for Better AI
byNeo4j
 
GraphTour London 2020 - Graphs for AI, Amy Hodler
byNeo4j
 
IRJET- Link Prediction in Social Networks
byIRJET Journal
 
Leveraging Graphs for Better AI
byNeo4j
 
GraphTour 2020 - Graphs & AI: A Path for Data Science
byNeo4j
 
How Graphs Enhance AI
byNeo4j
 
Transforming AI with Graphs: Real World Examples using Spark and Neo4j
byFred Madrid
 
Spark Summit EU talk by Reza Karimi
bySpark Summit
 

More from Neo4j

PDF
Google Cloud Presentation GraphSummit Melbourne 2024: Building Generative AI ...
byNeo4j
 
PDF
GraphRAG and Knowledge Graphs Exploring AI's Future
byNeo4j
 
PPTX
Neo4j Knowledge for Customer Experience.pptx
byNeo4j
 
PDF
GraphSummit Singapore Master Deck - May 20, 2025
byNeo4j
 
PPTX
Graphs & GraphRAG - Essential Ingredients for GenAI
byNeo4j
 
PDF
Master Deck: GraphSummit Bengaluru (Oct 7)
byNeo4j
 
PDF
ANZ Presentation: GraphSummit Melbourne 2024
byNeo4j
 
PDF
Hands-On GraphRAG Workshop: GraphSummit Melbourne 2024
byNeo4j
 
PDF
Telstra Presentation GraphSummit Melbourne: Optimising Business Outcomes with...
byNeo4j
 
PDF
Matinée GenAI & GraphRAG Paris - Décembre 24
byNeo4j
 
PDF
MASTERDECK GRAPHSUMMIT SYDNEY (Public).pdf
byNeo4j
 
PDF
Smarter Knowledge Graphs For Public Sector
byNeo4j
 
PDF
The Art of Possible - GraphTalk Paris Opening Session
byNeo4j
 
PPTX
GraphTalk New Zealand - The Art of The Possible.pptx
byNeo4j
 
PDF
Neo4j: The Art of the Possible with Graph
byNeo4j
 
PDF
Démonstration Supply Chain - GraphTalk Paris
byNeo4j
 
PDF
Swiss Life - Les graphes au service de la détection de fraude dans le domaine...
byNeo4j
 
PPTX
How Siemens bolstered supply chain resilience with graph-powered AI insights ...
byNeo4j
 
PDF
Jin Foo - Prospa GraphSummit Sydney Presentation.pdf
byNeo4j
 
PDF
Démonstration Digital Twin Building Wire Management
byNeo4j
 
Google Cloud Presentation GraphSummit Melbourne 2024: Building Generative AI ...
byNeo4j
 
GraphRAG and Knowledge Graphs Exploring AI's Future
byNeo4j
 
Neo4j Knowledge for Customer Experience.pptx
byNeo4j
 
GraphSummit Singapore Master Deck - May 20, 2025
byNeo4j
 
Graphs & GraphRAG - Essential Ingredients for GenAI
byNeo4j
 
Master Deck: GraphSummit Bengaluru (Oct 7)
byNeo4j
 
ANZ Presentation: GraphSummit Melbourne 2024
byNeo4j
 
Hands-On GraphRAG Workshop: GraphSummit Melbourne 2024
byNeo4j
 
Telstra Presentation GraphSummit Melbourne: Optimising Business Outcomes with...
byNeo4j
 
Matinée GenAI & GraphRAG Paris - Décembre 24
byNeo4j
 
MASTERDECK GRAPHSUMMIT SYDNEY (Public).pdf
byNeo4j
 
Smarter Knowledge Graphs For Public Sector
byNeo4j
 
The Art of Possible - GraphTalk Paris Opening Session
byNeo4j
 
GraphTalk New Zealand - The Art of The Possible.pptx
byNeo4j
 
Neo4j: The Art of the Possible with Graph
byNeo4j
 
Démonstration Supply Chain - GraphTalk Paris
byNeo4j
 
Swiss Life - Les graphes au service de la détection de fraude dans le domaine...
byNeo4j
 
How Siemens bolstered supply chain resilience with graph-powered AI insights ...
byNeo4j
 
Jin Foo - Prospa GraphSummit Sydney Presentation.pdf
byNeo4j
 
Démonstration Digital Twin Building Wire Management
byNeo4j
 

Recently uploaded

PDF
[DSC Europe 25] Andrzej Kowalczyk - AI - how to start small and grow in the f...
byDataScienceConferenc1
 
PPTX
Enhanced signature verification through the integration of divers' models, op...
bychWaqasZahid
 
PDF
Semantic Technology
byRathachai Chawuthai
 
PPTX
[DSC Europe 25] Marcos Heidemann - Beyond the Hype: Making AI Coding Assistan...
byDataScienceConferenc1
 
PDF
Operations and Data Support Portfolio | Delbert Ross Wanjala Investor Relatio...
byDelbertRossWanjala
 
PPTX
[DSC Europe 25] Borko Kozomora - Optimizing business workflows with advances ...
byDataScienceConferenc1
 
PPTX
[DSC Europe 25] Dubravko Culibrk - Deep Learning for Mammography.pptx
byDataScienceConferenc1
 
PPTX
[DSC Europe 25] Paula Garcia Esteban -Building the Future: The Role of Data S...
byDataScienceConferenc1
 
PPTX
[DSC Europe 25] Egor Krasheninnikov - The Control Stack: Building Guardrails ...
byDataScienceConferenc1
 
PDF
[DSC Europe 25] Tali Fulman - Guild Meetings, Then What? Building Data Commun...
byDataScienceConferenc1
 
PPTX
Artificial neural network Available for instant download in exchange for a bi...
byvinstonr
 
PDF
When Is the Right Time to Telegram Accounts_.pdf
by3tb3224qp2qbmz9iivnj
 
PPTX
[DSC Europe 25] Bojan Banjac - AI is always right when it comes to the matter...
byDataScienceConferenc1
 
PDF
Power BI Sales Analysis: Performance, Trends, and Strategic Insights my final...
bymaduruth09
 
PDF
How To Bu'y Verifi'ed kraken Account in US.pdf
byuerubi95
 
PDF
mythbusting-cdc-what-employers-need-to-know.pdf
byHenry Tapper
 
PPTX
National Art School Diploma
by897eemzixp
 
PPTX
REVIEWER-MddddddddddddATH-10-3rd-quarter.pptx
bydominicdaltoncaling2
 
PPTX
PPT sidang disertasi terbuka terbaru.pptx
byRifaiteguhSaputra
 
PDF
Torre Avianca- Modernity, Identity, and Technology in Colombian Architecture.pdf
bysanchezpipe1234
 
[DSC Europe 25] Andrzej Kowalczyk - AI - how to start small and grow in the f...
byDataScienceConferenc1
 
Enhanced signature verification through the integration of divers' models, op...
bychWaqasZahid
 
Semantic Technology
byRathachai Chawuthai
 
[DSC Europe 25] Marcos Heidemann - Beyond the Hype: Making AI Coding Assistan...
byDataScienceConferenc1
 
Operations and Data Support Portfolio | Delbert Ross Wanjala Investor Relatio...
byDelbertRossWanjala
 
[DSC Europe 25] Borko Kozomora - Optimizing business workflows with advances ...
byDataScienceConferenc1
 
[DSC Europe 25] Dubravko Culibrk - Deep Learning for Mammography.pptx
byDataScienceConferenc1
 
[DSC Europe 25] Paula Garcia Esteban -Building the Future: The Role of Data S...
byDataScienceConferenc1
 
[DSC Europe 25] Egor Krasheninnikov - The Control Stack: Building Guardrails ...
byDataScienceConferenc1
 
[DSC Europe 25] Tali Fulman - Guild Meetings, Then What? Building Data Commun...
byDataScienceConferenc1
 
Artificial neural network Available for instant download in exchange for a bi...
byvinstonr
 
When Is the Right Time to Telegram Accounts_.pdf
by3tb3224qp2qbmz9iivnj
 
[DSC Europe 25] Bojan Banjac - AI is always right when it comes to the matter...
byDataScienceConferenc1
 
Power BI Sales Analysis: Performance, Trends, and Strategic Insights my final...
bymaduruth09
 
How To Bu'y Verifi'ed kraken Account in US.pdf
byuerubi95
 
mythbusting-cdc-what-employers-need-to-know.pdf
byHenry Tapper
 
National Art School Diploma
by897eemzixp
 
REVIEWER-MddddddddddddATH-10-3rd-quarter.pptx
bydominicdaltoncaling2
 
PPT sidang disertasi terbuka terbaru.pptx
byRifaiteguhSaputra
 
Torre Avianca- Modernity, Identity, and Technology in Colombian Architecture.pdf
bysanchezpipe1234
 

Improving Machine Learning using Graph Algorithms

  • 1.
    Improve ML Predictionsusing Graph Algorithms Mark Needham, Neo4j Amy Hodler, Neo4j May 2019 #Neo4j #GraphAnalytics
  • 2.
    • Graphs forPredictions • Connected Features • Link Prediction • Neo4j + Spark Workflow Amy E. Hodler Graph Analytics & AI Program Manager, Neo4j Amy.Hodler@neo4j.com @amyhodler neo4j.com/ graph-algorithms-book Chapter 8: Graph + ML Spark & Neo4j Mark Needham Developer Relations Engineer, Neo4j Mark.needham@neo4j.com @markHneedham
  • 3.
    What in Commonis Predictive?
  • 4.
    Relationships: Strongest Predictors ofBehavior! “Increasingly we're learning that you can make better predictions about people by getting all the information from their friends and their friends’ friends than you can from the information you have about the person themselves” James Fowler David Burkus James Fowler Albert-Laszlo Barabasi
  • 5.
    Native Graph Platformsare Designed for Connected Data TRADITIONAL PLATFORMS BIG DATA TECHNOLOGY Store and retrieve data Aggregate and filter data Connections in data Real time storage & retrieval Real-Time Connected Insights Long running queries aggregation & filtering “Our Neo4j solution is literally thousands of times faster than the prior MySQL solution, with queries that require 10-100 times less code” Volker Pacher, Senior Developer Max # of hops ~3 Millions
  • 6.
    Graph Database Surgingin Popularity Trends since Jan 2013 DB-Engines.com
  • 8.
    Graph Data ScienceApplications
  • 9.
    • Current datascience models ignore network structure & complex relationships • Graphs add highly predictive features to existing ML models • Otherwise unattainable predictions based on relationships Novel & More Accurate Predictions with the Data You Already Have Machine Learning Pipeline
  • 11.
  • 12.
    Connection-related metrics aboutour graph, such as the number of relationships going into or out of nodes, a count of potential triangles, or neighbors in common. 14c What Are Connected Features?
  • 13.
    Query (e.g. Cypher) Real-time,local decisioning and pattern matching Graph Algorithms Libraries Global analysis and iterations You know what you’re looking for and making a decision You’re learning the overall structure of a network, updating data, and predicting Local Patterns Global Computation Deriving Connected Features
  • 14.
    Connected Feature Engineering FeatureEngineering is how we combine and process the data to create new, more meaningful features, such as clustering or connectivity metrics. Add More Descriptive Features: - Influence - Relationships - Communities Extraction
  • 15.
    17 Graph Feature Categories& Algorithms Pathfinding & Search Finds the optimal paths or evaluates route availability and quality Centrality / Importance Determines the importance of distinct nodes in the network Community Detection Detects group clustering or partition options Heuristic Link Prediction Estimates the likelihood of nodes forming a relationship Evaluates how alike nodes are Similarity Embeddings Learned representations of connectivity or topology
  • 16.
  • 17.
    19 Can we infernew interactions in the future? What unobserved facts we’re missing?
  • 18.
    + 50 yearsof biomedical data integrated in a knowledge graph Predicting new uses for drugs by using the graph structure to create features for link prediction Example: het.io
  • 19.
  • 20.
    Methods for LinkPrediction Algorithm Measures Run targeted algorithms and score outcomes Set a threshold value used to predict a link between nodes Machine Learning Use the measures as features to train an ML model Community Detection Link Prediction Similarity 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 2 4 15 1 3 4 7 12 1 5 6 1 1 0 Centrality
  • 21.
  • 22.
    • Citation NetworkDataset - Research Dataset – “ArnetMiner: Extraction and Mining of Academic Social Networks”, by J. Tang et al – Used a subset with 52K papers, 80K authors, 140K author relationships and 29K citation relationships • Neo4j – Create a co-authorship graph and connected feature engineering • Spark and MLlib – Train and test our model using a random forest classifier 24 Predicting Collaboration with a Graph Enhanced ML Model
  • 23.
    Our Link PredictionWorkflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize
  • 25.
    Our Link PredictionWorkflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identified sparse feature areas Feature Engineering: New graphy features
  • 26.
    Graph Algorithms Usedfor Feature Engineering (few examples) Preferential Attachment measure the closeness of nodes based on shared neighbors Common Neighbors measures the number of possible neighbors (triadic closure) Illustration be.amazd.com/link-prediction/
  • 27.
    Graph Algorithms Usedfor Feature Engineering (few examples) Triangle counting and clustering coefficients measure the density of connections around nodes Louvain Modularity identifies interacting communities and hierarchies
  • 28.
    Our Link PredictionWorkflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identified sparse feature areas Feature Engineering: New graphy features Train / Test Split Resample: Downsampled for proportional representation
  • 29.
  • 30.
    32 Test/Train Split 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 24 15 1 3 4 7 12 1 5 6 1 1 0 2 12 3 3 0 4 9 4 8 1 7 10 12 36 1 8 11 2 3 0
  • 31.
    33 Test/Train Split 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 24 15 1 3 4 7 12 1 5 6 1 1 0 2 12 3 3 0 4 9 4 8 1 7 10 12 36 1 8 11 2 3 0 Train Test
  • 32.
    OMG I’m Good! DataLeakage! Graph metric computation for the train set touches data from the test set. Did you get really high accuracy on your first run without tuning?
  • 33.
    Train and TestGraphs: Time Based Split 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 2 4 15 1 3 4 7 12 1 5 6 1 1 0 Train Test 1st Node 2nd Node Common Neighbors Preferential Attachment label 2 12 3 3 0 4 9 4 8 1 7 10 12 36 1 < 2006 >= 2006
  • 34.
    Train and TestGraphs: Time Based Split 1st Node 2nd Node Common Neighbors Preferential Attachment label 1 2 4 15 1 3 4 7 12 1 5 6 1 1 0 Train Test 1st Node 2nd Node Common Neighbors Preferential Attachment label 2 12 3 3 0 4 9 4 8 1 7 10 12 36 1
  • 35.
  • 36.
    There are significantlymore negative examples than positive ones: # negative examples = (# nodes)² - (# relationships) - (# nodes) 38 Class Imbalance
  • 37.
    A very highaccuracy model could predict that a pair of nodes are not linked. 39 Class Imbalance
  • 38.
  • 39.
    Our Link PredictionWorkflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identified sparse feature areas Feature Engineering: New graphy features Train / Test Split Resample: Downsampled for proportional representation Model Selection: Random Forest Ensemble method
  • 40.
  • 41.
    Training Our Model Thisis one decision tree in our Random Forest used as a binary classifier to learn how to classify a pair: predicting either linked or not linked.
  • 42.
    4 Models Trained withMultiple Graph Features Graph Features: • Common Authors “Graphy” Model Common Authors Model Triangles Model Community Model Graph Features: • Preferential Attachment • Total Neighbors Graph Features: • Min & Max Triangles • Min & Max Clustering Coefficient Graph Features: • Label Propagation • Louvain Modularity
  • 43.
    Our Link PredictionWorkflow Import Data Create Co-Author Graph Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identified sparse feature areas Feature Engineering: New graphy features Train / Test Split Resample: Downsampled for proportional representation Precision, Accuracy, Recall ROC Curve & AUC Model Selection: Random Forest Ensemble method
  • 44.
    Measures Accuracy Proportion oftotal correct predictions. Beware of skewed data! Precision Proportion of positive predictions that are correct. Low score = more false positives Recall / True Positive Rate Proportion of actual positives that are correct. Low score = more false negatives False Positive Rate Proportion of incorrect positives ROC Curve & AUC X-Y Chart mapping above 2 metrics (TPR and FPR) with area under curve
  • 45.
    Result: First ModelROC & AUC Problematic False Positives! Common Authors Model 1
  • 46.
    Result: All ModelsCommon Authors Model 1 Community Model 4
  • 47.
    Iteration & Tuning:Feature Influence For feature importance, the Spark random forest averages the reduction in impurity across all trees in the forest Feature rankings are in comparison to the group of features evaluated Also try PageRank! Try removing different features (LabelPropagation)
  • 48.
    Graph Machine LearningWorkflow Data aggregation Create and store graphs Extract Data & Store as Graph Explore, Clean, Modify Prepare for Machine Learning Train Models Evaluate Results Productionize Identify uninteresting features Cleanse (outliers+) Feature engineering/ extraction Train / Test split Resample for meaningful representation (proportional, etc.) Precision, accuracy, recall (ROC curve & AUC) SME Review Cross-validation Model & variable selection Hyperparameter tuning Ensemble methods
  • 49.
    Resources • neo4j.com/sandbox • neo4j.com/developer/ graph-algorithms/ •community.neo4j.com Data & Code: • This example from O’Reilly book bit.ly/2FPgGVV (ML Folder) Amy.Hodler@neo4j.com @amyhodler neo4j.com/ graph-algorithms-book
  • 50.
  • 51.
    53 Connected Feature Extraction FeatureExtraction is how when we change the shape or format of the data to be usable in a machine learning pipeline. For example, from a graph, we extract the relevant subset of the data into a tabular format for model building.
  • 52.
    Connected Feature Selection FeatureSelection is how we reduce the number of features used in a model to a relevant subset. This can be done algorithmically or based on domain expertise, but the objective is to maximize the predictive power of your model while minimizing overfitting.
  • 53.
    720+ 7/10 12/2 5 8/10 53K+ 100+ 300+ 450+ Adoption Top Retail Firms TopFinancial Firms Top Software Vendors Customers Partners • Creator of the Neo4j Graph Platform • ~250 employees • HQ in Silicon Valley, other offices include London, Munich, Paris and Malmö Sweden • $80M new funding led by Morgan Stanley & One Peak. Total $160M from Fidelity, Sunstone, Conor, Creandum, and Greenbridge Capital • Over 15M+ downloads & container pulls • 325+ enterprise subscription customers with over half with >$1B in revenue Ecosystem Startups in program Enterprise customers Partners Meet up members Events per year Industry’s Largest Dedicated Investment in Graphs Neo4j - The Graph Company
  • 54.
    Strictly ConfidentialStrictly Confidential 56 HelpingThe World To Make Sense of Data ICIJ used Neo4j to uncover the world’s largest journalistic leak to date, The Panama Papers NASA uses Neo4j for a “Lessons Learned” database to improve effectiveness in search missions in space Neo4j is used to graph the human body, map correlations, identify cause & effect and search for the cure for cancer SAVING DEMOCRACY MISSION TO MARS CURING CANCER
  • 55.
    Graph and MLAlgorithms in Neo4j • Parallel Breadth First Search & DFS • Shortest Path • Single-Source Shortest Path • All Pairs Shortest Path • Minimum Spanning Tree • A* Shortest Path • Yen’s K Shortest Path • K-Spanning Tree (MST) • Random Walk • Degree Centrality • Closeness Centrality • CC Variations: Harmonic, Dangalchev, Wasserman & Faust • Betweenness Centrality • Approximate Betweenness Centrality • PageRank • Personalized PageRank • ArticleRank • Eigenvector Centrality • Triangle Count • Clustering Coefficients • Connected Components (Union Find) • Strongly Connected Components • Label Propagation • Louvain Modularity – 1 Step & Multi-Step • Balanced Triad (identification) • Euclidean Distance • Cosine Similarity • Jaccard Similarity • Overlap Similarity • Pearson Similarity Pathfinding & Search Centrality / Importance Community Detection Similarity neo4j.com/docs/ graph-algorithms/current/ Updated April 2019 Link Prediction • Adamic Adar • Common Neighbors • Preferential Attachment • Resource Allocations • Same Community • Total Neighbors
  • 56.
    Conceive Code Compute Store Non-Native Graph DBNativeGraph DB RDBM S Optimized for graph workloads Connectedness Differentiates Neo4j
  • 57.
    Neo4j is anenterprise-grade native graph platform that enables you to: • Store, reveal and query data relationships • Traverse and analyze any levels of depth in real-time • Add context and connect new data on the fly 59 Who We Are: Leader in Graph Innovations • Performance • ACID Transactions • Schema-free Agility • Graph Algorithms Designed, built and tested natively for graphs from the start for: • Developer Productivity • Hardware Efficiency • Global Scale • Graph Adoption Graph Transactions Graph Analytics Data Integration Development & Admin Analytics Tooling Drivers & APIs Discovery & Visualization
  • 58.
    60 • Record “CyberMonday” sales • About 35M daily transactions • Each transaction is 3-22 hops • Queries executed in 4ms or less • Replaced IBM Websphere commerce • 300M pricing operations per day • 10x transaction throughput on half the hardware compared to Oracle • Replaced Oracle database • Large postal service with over 500k employees • Neo4j routes 7M+ packages daily at peak, with peaks of 5,000+ routing operations per second. Handling Large Graph Work Loads for Enterprises Real-time promotion recommendations Marriott’s Real-time Pricing Engine Handling Package Routing in Real-Time
  • 59.
    Recommendations Dynamic PricingIoT-applicationsFraud Detection Real-Time Transaction Applications Generate and Protect Revenue Customer Engagement Metadata and Advanced Analytics Data Lake Integration Knowledge Graphs for AI Risk Mitigation Generate Actionable Insights Network Management Supply Chain Efficiency Identity and Access Management Internal Business Processes Improve Efficiency and Cut Costs Graph Use Cases by Value Proposition
  • 60.
    Softwar e Financial Services Teleco m Retail & ConsumerGoods Media & Entertainment Other Industries Airbus 62 Copyright © 2017 Neo4j, Inc. Company Confidential
  • 61.
    Graph Transactions Graph Analytics Data Integration Development & Admin Analytics Tooling Drivers& APIs Discovery & Visualization Developers Admins Applications Business Users Data Analysts Data Scientists Enterprise Data Hub Native Graph Platform: Tools for Many Users
  • 62.
    Collections-Focused Multi-Model, Documents, Columns &Simple Tables, Joins Neo4j is designed for data relationships Different Paradigms NoSQL Relational DBMS Neo4j Graph Platform Connections-Focused Focused on Data Relationships Development Benefits Easy model maintenance Easy query Deployment Benefits Ultra high performance Minimal resource usage
  • 63.
    How Neo4j Fits— Common Architecture Patterns From Disparate Silos To Cross-Silo Connections From Tabular Data To Connected Data From Data Lake Analytics to Real-Time Operations
  • 64.
    Cypher: Powerful &Expressive Query Language MATCH (:Person { name:“Dan”} ) -[:MARRIED_TO]-> (spouse) MARRIED_TO Dan Ann NODE RELATIONSHIP TYPE LABEL PROPERTY VARIABLE
  • 65.
    Neo4j Bloom 67 • Highfidelity • Scene navigation • Property views • Search suggestions • Saved phrase history • Property editor • Schema perspectives • Bloom chart type • Visualize • Communicate • Discover • Navigate • Isolate • Edit • Share
  • 66.
    68 Real-Time Recommendations Fraud Detection Network & IT Operations MasterData Management Knowledge Graph Identity & Access Management Common Graph Technology Use Cases AirBnb
  • 67.
    Graphs Drive Innovation 69 ContextPaths Auto-Graphs Graph Layers 1st Graph Cross- Connect Cross-tech applications Internet of Things operations Transparent Neural Networks Blockchain-managed systems Adjacent graph layers inspire new innovations Metadata / Risk Management Knowledge Graphs AI- Powered Customer Experiences Connect unlike objects such as people to products, locations Mobile app explosion Recommendation engines Fraud detectors Desire for more context to follow connections Connects like objects People, computer networks, telco, etc
  • 68.
    Business Problem • Findrelationships between people, accounts, shell companies and offshore accounts • Journalists are non-technical • Biggest “Snowden-Style” document leak ever; 11.5 million documents, 2.6TB of data Solution and Benefits • Pulitzer Prize winning investigation resulted in robust coverage of fraud and corruption • PM of Iceland & Pakistan resigned, exposed Putin, Prime Ministers, gangsters, celebrities (Messi) • Led to assassination of journalist in Malta Background • International Consortium of Investigative Journalists (ICIJ), small team of data journalists • International investigative team specializing in cross-border crime, corruption and accountability of power • Works regularly with leaks and large datasets ICIJ Panama Papers INVESTIGATIVE JOURNALISM Fraud Detection / Knowledge Graph70
  • 69.
    Thomson Reuters Graph 71 •Data Fusion for Portfolio Managers • Graph layers
  • 70.
    Background • Personal shoppingassistant • Converses with buyer via text, picture and voice to provide real-time recommendations • Combines AI and natural language understanding (NLU) in Neo4j Knowledge Graph • First of many apps in eBay's AI Platform Business Problem • Improve personal context in online shopping • Transform buyer-provided context into ideal purchase recommendations over social platforms • "Feels like talking to a friend" Solution and Benefits • 3 developers, 8M nodes, 20M relationships • Needed high-performance traversals to respond to live customer requests • Easy to train new algorithms and grow model • Generating revenue since launch eBay for Google Assistant ONLINE RETAIL Knowledge Graph powers Real-Time Recommendations72 EE Customer since 2016 Q3
  • 71.
    Background • Over 7Mcitizens suffer from Diabetes • Connecting over 400 researchers • Incorporates over 50 databases, 100k’s of Excel workbooks, 30 database of biological samples • Sought to examine disease from as many angles as possible. Business Problem • Genes are connected by proteins or to metabolites, and patients are connected with their diets, etc… • Needed to improve the utilization of immensely technical data • Needed to cater to doctors and researchers with simple navigation, communication and connections of the graph. Solution and Benefits • Dr. Alexander Jarasch, Head of Bioinformatics and Data Management • Scientists can conduct parallel research without asking the same questions or repeating tests • Built views like a liver sample knowledge graph DZD - German Center for Diabetes Research Medical Genomic Research73 EE Customer since 2016 Q4