Oracle Corporation
07/26/2024 | Press release | Archived content
Empowering the banking sector with OCI Infra AI and ML solutions
The banking sector faces an increasingly complex landscape that requires robust, scalable, and secure technology solutions. Oracle Cloud Infrastructure (OCI) offers a powerful environment specifically tailored for artificial intelligence (AI) and machine learning (ML) projects, which can significantly enhance various banking operations. This blog post explores the advantages of using OCI for AI and ML projects in banking, highlighted by a practical use case involving fraud detection.
The banking industry has already begun using OCI Generative AI services for fraud detection. The approach was chosen for its robust security features, high-performance computing, seamless data integration, scalability, and Oracle's proven expertise in financial services, ensuring compliance and efficiency in detecting and preventing fraud.
Model training and inferencing with OCI
Model training and inferencing with OCI
Why OCI is ideal for AI and ML in banking
OCI offers the following features and benefits for the banking industry:
Security and compliance: OCI provides an array of security features that comply with global and regional regulations, crucial for the banking sector. Examples include data encryption at rest and in transit, network isolation, and comprehensive identity and access management.
High-performance computing (HPC): For AI and ML workloads that require processing large volumes of data, OCI offers HPC options, including GPU and CPU instances that can scale according to demand. A key differentiator for OCI is its advanced clustering capabilities, which enable efficient parallel processing and improved performance for complex computations, as illustrated in the accompanying diagram. This addition highlights the clustering capabilities of OCI, aligning with the information provided in the diagram.
Data integration and management: OCI supports seamless data integration and management, which is vital for banks dealing with heterogeneous data sources. This integration capability ensures that data remains consistent, accurate, and readily available.
Scalability and flexibility: Banks can scale their resources up or down on OCI without significant upfront investments, allowing them to adjust to changing market conditions and customer demands effectively.
Proven expertise in financial services: Oracle has a long-standing presence in the financial services industry, offering tailored solutions that address specific sector needs from risk management to customer analytics.
Use case: AI-driven fraud detection system
Architecture diagram for the example use case.
In this example use case, a multinational bank struggles with the increasing sophistication of financial fraud, affecting customer trust and incurring significant losses. The bank needs an advanced solution to detect and prevent fraudulent transactions in real time, minimizing false positives that can disrupt genuine transactions.
As a solution, they implement an AI-driven fraud detection system using OCI using a both online and offline approach. Online and offline processing are essential to meet business requirements for transaction approval speed and detailed analysis. Online processing allows for real-time transaction approval or rejection, ensuring swift customer service. Offline processing enables in-depth analysis of transactions at a later time, improving fraud detection accuracy and refining models without impacting real-time operations. The two processing boxes represent these distinct yet complementary stages.
Online and offline processing use different models. Online processing requires a streamlined, real-time model optimized for speed to quickly approve or reject transactions. Offline processing, on the other hand, utilizes a more complex, detailed model for thorough analysis and deep dives into transaction data, allowing for continuous improvement and adaptation of the fraud detection system. The online processing model is typically called the "transactional model," while the offline processing model is referred to as the "analytical model."
Implementation steps
To implement their solution, the bank uses the following steps:
Data aggregation: Collect transactional data across different channels, such as online, mobile, and ATMs.
Secure storage and processing: Store sensitive data securely on OCI, utilizing Oracle's autonomous databases for high performance and built-in security features.
Workflow for the example secure storage and processing, from data sources to machine learning in Oracle Database to models at work to the results.
Workflow for the example secure storage and processing, from data sources to machine learning in Oracle Database to models at work to the results.
Model development and training: Use OCI Data Science services to develop and train ML models to identify potential fraud patterns. Models are trained using historical fraud data alongside normal transactions to understand and predict fraudulent behavior accurately.
The bank monitors drift and implements retraining by continuously analyzing new transactions and adapting the models to evolving fraud tactics. They achieve this goal through OCI Data Science, which facilitates ongoing learning from real-time data, ensuring that the models remain accurate and effective over time.
Workflow for developing and trading models, from data sources to data persistence to data access to the final result
Workflow for developing and trading models, from data sources to data persistence to data access to the final result.
Real-time analysis and deployment: Deploy the models on OCI, integrating them with the transaction processing systems to analyze transactions in real-time.
Continuous learning: Models continuously learn from new transactions and adapt to evolving fraud tactics, improving their accuracy over time.
As a result, the bank reduces fraudulent transactions within the first year of implementation, while improving the accuracy of real-time fraud detection.
The bank used the OCI Streaming service for online or mobile data to handle real-time data processing. Most of the transactional data processed is in real time. They used Oracle GoldenGate for data aggregation across different channels, ensuring seamless integration and management of diverse data sources.
OCI Data Science features used
This solution uses the following OCI Data Science features:
Model development and training
Historical data analysis
Continuous learning and adaptation
Real-time fraud detection integration
Data aggregation and preprocessing
Deployment of ML models
The Autonomous Database service's built-in machine learning (ML) features helped to develop and train the ML models for fraud detection, utilizing the following Autonomous Database high-performance capabilities and built-in security features:
Built-in ML features
High-performance capabilities
Secure data storage and processing
Autonomous data management and integration
Real-time data analysis capabilities
How OCI made a difference
OCI's services contributed to the success of the deployment with the following factors:
Performance: OCI's high-performance compute instances expedited the data processing and model training phases, enabling real-time fraud detection.
Security: OCI's comprehensive security features ensured that all data, a critical concern in banking, remained secure against threats.
Scalability: The bank could effortlessly scale its operations to handle peaks in transaction volumes, especially during high-traffic periods.
If any bank implements this solution with described Oracle Technologies, the following table provides a clear and quantifiable overview of the benefits and improvements offered by utilizing OCI for AI and ML solutions in the banking sector:
Metric
Value or description
Explanation
Fraud reduction (%)
20% reduction in fraudulent transactions in the first year
The implementation of AI-driven fraud detection resulted in a 20% decrease in fraudulent transactions.
False positive reduction (%)
15% reduction in false positives
AI models improved detection accuracy, reducing false positives by 15%.
Data encryption (in use)
100%
All data is encrypted at rest and in transit, ensuring maximum security.
Scalability (Instances)
Scalable from 1 to thousands of GPU and CPU instances
OCI offers flexibility to scale computing resources as needed.
Data processing speed (Improvement)
Two-times faster processing speed
HPC options resulted in data processing being twice as fast.
Integration time (Reduction)
30% reduction in integration time
Seamless data integration capabilities reduced the time needed for integrating heterogeneous data sources by 30%.
Security compliance
100% compliance with global and regional regulations
OCI ensures full compliance with necessary regulations, crucial for the banking sector.
Initial investment (Reduction)
50% reduction in upfront investment costs
Flexibility to scale without significant upfront investments led to a 50% reduction in initial costs.
Model training time (Reduction)
40% reduction in model training time
Use of high-performance Compute instances and OCI Data Science reduced model training time by 40%.
Transaction volume handling (Improvement)
Capable of handling two times the peak transaction volumes
The system can scale to manage double the peak transaction volumes compared to previous capabilities.
Using the following Oracle products and services can help you achieve these results:
OCI
OCI Data Science
OCI Streaming
Oracle Autonomous Database
Oracle GoldenGate
OCI high-performance Compute instances
OCI Data Integration and Management
Conclusion
OCI provides a secure, scalable, and efficient platform that's ideal for deploying sophisticated AI and ML solutions in the banking sector. By using OCI, banks can enhance their ability to combat fraud, improve customer service, and optimize operations, all while adhering to strict regulatory requirements.
This post provides an insight into how banks can use OCI to foster innovation and efficiency in the banking sector, particularly using AI and ML technologies, driving forward a new era of digital banking solutions.
For more information on Oracle Cloud Infrastructure's capabilities and how it supports the banking industry, explore the following resources:
Oracle Cloud Infrastructure for financial services
OCI's approach to AI and machine learning
Oracle Cloud Infrastructure Security Architecture
Introduction to OCI Data Science
Ravindra Oza
Technical Program Manager
Multi Cloud Strategic Leader-IIMA-Leadership Skills-PMP-ITIL-IBM Design Thinker -GCP|OCI|AWS|Azure (IaaS,PaaS,DBaaS) ,Exadata DMA, Oracle Database Administration & GG Expert