Skip to content 
GrandhooWuji Intelligent Risk Identification Solution
Overview
As credit business advances, banks need to conduct real-time monitoring and analysis of a large volume of data to identify fraudulent activities and enhance the accuracy of risk identification. The “Wuji Intelligent Risk Identification Solution”, through core functions such as data processing, marketplace construction, and data modeling, seamlessly addresses the comprehensive demand for data analysis in the credit business of the banking industry and provides guarantee and support for the entire credit business process through applications of data analysis. Through real-time detection of incremental business data and identification of fraud based on the risk rule base, “Wuji Intelligent Risk Identification Solution” is able to better evaluate business development and provide decision support for refined management. By real-time monitoring of incremental business data and identifying fraudulent activities based on a library of risk rules, the Wujie Intelligent Risk Identification Solution is able to better evaluate business development and provide decisive support for refined management.
“Wuji Intelligent Risk Identification Solution” is based on an original graph computing rule engine, which automatically analyzes the knowledge network constructed by indivudal information, customer behavior, transaction data, credit information, and other informative data and presents the results in the form of graph-pattern association rules. Through the applications of rules of the incremental data, it effectively monitors transaction behavior in real time, identifies abnormal behavior, establishes a risk rule base, and empowers financial institutions with risk control scenarios.
Challenges
Single dimension of risk assessment
Under traditional risk management methods, retail banking primarily relies on credit data for credit risk assessment. The limited data demension and the critical issue that the vast majority of individuals lacking credit history face obstacles in accessing formal financial services are the primary challenges of the business.
Inefficient manual approvals
The traditional retail credit risk control of commercial banks mainly relies on manual approval, resulting in less timeliness and accuracy than advanced algorithmic models. Therefore, banks urgently need to solve this problem.
Risk control rules are not comprehensive
Currently, the risk control rules are mainly derived from the basic information and behavioral characteristics attributes of individuals, and lack the structural characteristics attributes of relationship networks.
Architecture
Benefits
Risk rules are automatically generated for more efficient rule entry
Based on user data and business data, the “Wuji Intelligent Risk Identification Solution” constructs the graph data structure and automatically discovers patterns of fraudulent behavior through the self-developed graph computing engine. Additionally, it is capable of generating rules, which possess high interpretability and tracability, based on the fraudulent behavior. After practical testing, graph computing has demonstrated a performance improvement of over 50 times compared to traditional machine learning training, saving significant manpower and time costs, thereby enhancing anti-fraud capabilities for financial institutions
Flexible rule configuration and real-time risk prediction
After analyzing the rules generated from fraudulent behavior, a risk rule library will be established for rule management. This supports rule classification and the creation of rule sets. It also supports the creation, viewing, copying, modification and deletion of rules. Additionally, the system allows the import of manually crafted fraud rules to enrich the risk rule library, providing support for subsequent fraud detection.Based on the risk rule library, real-time detection of incremental business data is conducted to promptly identify fraudulent activities。
Combining graph relational features for more comprehensive rule mining
Combining network relationships between multiple entities, user transaction network relationships, and user-account-device network topology relationships to capture network structured features is more conducive to identifying fraud rings.