Graph-based text data relationship & semantic analysis

Detect fraudulent data via graph pattern analysis

Identify unknown data types with graph clustering

Graph inference for highly accurate prediction based on training data

**Use Graph Analytics for Big Data!**

**G-PAS** (Graph-based Predictive Analytics System) uses a property graph data model to store big data.

**Real Data**

**RDB Table**

**Property Graph**

**Table -> Cost incurs when processing graph analysis**

(During converting & mathematical algorithm operation)

**Save data in Graph**

**Achieve low cost & high efficiency**

*Property Graph Data Model:

A standard model commonly used in most graph databases around the world, including AgensGraph

**Real Data**

**RDB Table**

**Table -> Cost incurs when processing graph analysis**

(During converting & mathematical algorithm operation)

**Property Graph**

**Save data in Graph
Achieve low cost & high efficiency**

*Property Graph Data Model:

A standard model commonly used in most graph databases around the world, including AgensGraph

**Features and Differences**

#### G-PAS helps existing systems overcome limitations and fulfill requirements and needs

**Requirements**

**Existing Limitations**

**Existing Limitations**

### Predictive Analysis with Unstructured Data

Incurs wasted storage space for storing unstructured data (lower storage efficiency)

**Utilize schemaless property of graph data structure to effectively store unstructured data**

Incurs wasted storage space for storing unstructured data (lower storage efficiency)

**Utilize schemaless property of graph data structure to effectively store unstructured data**

**Real-time data** prediction/analysis

Data relationship analysis is unavailable due to the real-time domain being set separately from transaction management

***Graph DB performs transaction management and analyzes relationships between data in graph form**

Data relationship analysis is unavailable due to the real-time domain being set separately from transaction management

***Graph DB performs transaction management and analyzes relationships between data in graph form**

### Knowledge DB for **complex data** management

Cannot express the relationship between knowledge, only a simple keyword-based knowledge search system is implemented

**Able to deduct and derive information from the relationship between knowledge**

Cannot express the relationship between knowledge, only a simple keyword-based knowledge search system is implemented

**Able to deduct and derive information from the relationship between knowledge**

*Transaction: a sequence of information exchanged to change the state of the database

**Graph Analysis Method**

##### Bitnine added years of experience in its unique analysis technology and developed specialized services.

**Text data relationship & semantic analysis**

**Graph-based **

**Text data relationship & semantic analysis**

Understand how trend changes over time by expressing a graph of keyword appearance patterns within vast text data

Core analysis techniques: TF-IDF, LDA Topic Modeling, etc

**TF-IDF Text network**

Quantitatively measures the importance of a specific word in a document (TF-IDF) and by creating a graph, it is possible to identify the trend of the words in a topic close to each other

**LDA Topic modeling**

One of the probabilistic topic modeling techniques that generate a knowledge graph by estimating the distribution of words and documents by topic

**Graph pattern analysis**

**Detect fraudulent data via**

**graph pattern analysis**

Fraudulent data can be detected in real-time by analyzing the trends of patterns in the graph data.

Core analysis techniques: Motif analysis, Path analysis, etc.

**Path Analysis**

The overall analysis of the path within the graph. Includes analysis leading to optimization of the overall graph flow to find the shortest distance between two nodes.

Motif Analysis

Statistically analyzes the frequency of all existing connection patterns in graph data. Helps understand the rules of each entity.

**Graph Clustering**

**Identify unknown data types with**

**Graph Clustering**

Based on clusters of graph data, new types of data previously unclassified in the existing system can be identified

Core analysis techniques : Graph similarity, Community Detection, etc.

**Graph Similarity**

Define similarity by numerically quantifying the connected structure of graph data

**Community Detection**

A method of dividing clusters based on the density of connection patterns in the graph.

**Graph inference**

**Graph inference**** for highly accurate**

**prediction based on training data**

Learn complex domain data and implement highly accurate qualitative/quantitative data inference with either bayesian network or graph-based deep learning technology

Core analysis techniques : Graph-based Deep Learning, Bayesian Network, etc.

**Graph Neural Network**

A deep learning technique that can recieve graph data as an input value. GNN(Graph Neural Network), GCN(Graph Convolutional Network) are two main methods.

**Bayesian Network**

Probabilistic inference of result values according to various conditions. Use graph as an operator

**Graph Analytics Use Case**

#### Introducing use case of Graph-based Predictive Analytics System in big data environment

**Text data relationship & semantic analysis**

**Graph-based**

**text data relationship & semantic analysis**

**Housing Policy Analysis System**

Developed NLP-based predictive analytics system to perform quantitative/qualitative analysis on policy effect. Various data (i.e. press articles, local government press releases, housing price data, etc.) are gathered and expressed in graph data to analyze the effect of housing policy.

**Social Media Analysis System**

Developed an analysis solution that can filter complex social media environments, identify and analyze important themes and ideas of one’s online activities (ex. news on a specific topic or finding new influencers, identifying links shared by followers or friends, etc.)

**Graph pattern analysis**

**Detect fraudulent data via**

**graph pattern analysis**

**Traveler Information Analysis System**

In order to prevent crimes such as smuggling/terrorism, it was necessary to find the potential offender and their hidden companions among the numerous immigration travelers within a short period of time. By using graph DB, a real-time social network analysis system was implemented and the smugglers were identified and captured.

**Real-time Fraud Detection System**

The fraudulent transaction pattern previously undetected by a rule-based FDS was successfully detected by Graph based Fraud Detection System. Transaction history and relationship was defined by graph modeling and fake deposit was detected via graph analysis and pattern search algorithm.

**Graph clustering**

**Identify unknown data types with**

**Graph clustering**

**Cyber Threat Intelligence**

Established cluster classifying integrated analysis system for associations between many types of data collected from whitelist/blacklist, security vulnerabilities, malicious codes, and news and data, etc.

**Real-time Recommendation System based on Pattern Similarity of Card Usage**

Pattern inquiry query is performed in a DB to calculate similarity of card usage patterns. A graph environment is built to see customer similarity according to card usage pattern cluster analysis. Thus, recommendations for individual became available and hidden patterns not found in the existing system was discovered.

**Graph inference**

**Graph inference **** for highly accurate**

prediction based on training data

prediction based on training data

**AI-based Insulation Material Prescription System**

Many trials and errors occurred when developing a new synthesis of inorganic compounds. When all data is quantified in graph data and AI technology is applied, a new product prescription method was developed.

**Financial Regulation Terms & Conditions Analysis System**

Interpretation of complex financial regulation laws is handled manually every time and due to rising social costs, all laws and regulations are loaded in a hierarchical graph structure to realize the hierarchical relationship within the regulatory law.

** Solution**

#### Introducing Product and Service

ProductGraph PAS Product

** : Graph Predictive Analytics engine**

Graph data-based big data predictive analytics engine

** : Graph Data Storage**

Efficiently stores and manages graph data

** : Graph Data Analysis Tool**

Analyze and visualize graph data with maximized convenience

ServiceGraph PAS Service

**Data Analytics Consulting**

Data-based service design and advanced master plan consulting

**Custom Analytics Algorithm Development**

Customize analytics algorithms specialized for domain data

**Graph Data Analytics Education**

Specialized education for graph data analysis based on data and purpose