GRAPH ANALYSIS AND INFERENCE
Overview
The graph is a knowledge base of graph data. It integrates multiple types of data and extracts information through the connection of points and edges. Compared with traditional databases, it can present the relationship between information in a larger, complete, and faster manner. Trend analysis of Gartner pointed out that the proportion of graph analysis in enterprise applications will increase from 10% in 2021 to 80% in 2025. Through the graph, both the correlation analysis and inference capabilities of AI and machine learning can be improved. It is suitable for applications that require a huge amount of complex correlation operations, such as recommendation systems, search, customer management, social media analysis, medical disease tracking, technology law enforcement, financial fraud detection, supply chain traceability management, and other fields.GRAPH ANALYSIS AND INFERENCE
CORE TECHNOLOGY
- Graph Feature Extraction and Knowledge Fusion
- Graph Structure Algorithm
- Graph Neural Network Learning
Graph Construction and Application Process
GRAPH ANALYSIS AND INFERENCE
Application Status
- CHT-TL focuses on the research and development of graph-related technologies, including graph construction, establishment and management of qualitative/quantitative indicators, feature extraction and fusion, graph analysis and inference technology based on specific applications, etc. Our graph-related technologies can assist enterprises in quickly establishing related applications and solutions.
- The video graph has been built and imported into the Chunghwa Telecom's Hami video and video search and recommendation applications of MOD. The video recommendation technology contributes to increasing both the average monthly clicks by 47% and the number of watched videos by 3.75 times. The result shows that the graph is more suitable for product recommendations, and diversified correlation features also help increase product exposure.
- In terms of user behavior analysis, comprehensive user profile analysis can be provided through behavioral graph analysis. And the graph community detection algorithm can obtain more accurate user groups, which can be used to create personalized services or marketing plans, which are beneficial to customer management, marketing, and advertising matching.
- For applications in specific fields, through graph analysis and inference, it is possible to gain insights into abnormal information or behaviors. And these technologies are also able to dig out potential abnormal patterns or undisclosed risk events, which is conducive to risk management, such as fraud detection applications.