Countries as diverse as Portugal, Guatemala, Chile, Colombia, Mexico, and Oman have been the focus of our Proofs of Concept in recent months, and the results for our artificial intelligence engine, COCO, have been more than positively surprising. Training with transactional data from such different markets strengthens its responsiveness and sharpens its understanding of banking data.
Óscar Barba
Co-founder & CTO of Coinscrap Finance
Smart adaptation to diverse financial markets
One of the biggest challenges in expanding financial technologies is adapting to environments with different regulations, banking systems, and consumption habits. However, transfer learning allows us to leverage knowledge gained in one market and apply it efficiently in another, reducing implementation times and improving the accuracy of transactional analysis.
This approach offers key advantages for banks seeking to scale their technology globally. To achieve a quick implementation in new countries, there is no need to start from scratch, as transfer learning allows them to expand naturally. By making the right adjustments, total adaptation to each region’s specific needs is achieved.
One of the competitive advantages of artificial intelligence is cost and time optimization. The effort required to train models in each market is reduced, and information processing time is optimized. COCO can analyze and enrich banking transactional data in milliseconds and offer an accuracy of over 90%.
💡 Did you know that…
The ability to quickly and accurately transfer financial solutions to different markets is redefining the global banking landscape, offering more efficient and personalized services for users worldwide.
Technical aspects of transfer learning in transactional data
Transfer learning in the context of banking transactional data presents a series of technical challenges and opportunities. COCO has been designed to learn financial patterns from a dataset and transfer that knowledge to new environments. Below are the most important technical aspects when working with an AI engine:
Data normalization
Different banks and countries manage different transactional categories. Normalization and standardization are crucial to ensure consistency in learning.
Multimodal learning
COCO not only analyzes transactions but also spending patterns. The ability to integrate different data sources enhances the robustness of our model.
Hyperparameter tuning
Since each market has unique characteristics, we use hyperparameter tuning techniques to optimize AI accuracy in each environment.
Bias mitigation
Bias in financial data can lead to incorrect predictions. We apply balancing and calibration strategies to ensure fairness and representativeness in recommendations.
Security and regulatory compliance:
When working with sensitive data, COCO is designed to operate under strict security protocols and in compliance with regulations like GDPR, PSD2, and ISO27001.
COCO: An ever-evolving AI engine
The work done with banks in diverse markets like Portugal, Guatemala, Chile, Colombia, Mexico, and Oman has been crucial in improving our technology. With each project, COCO has learned to categorize transactions more accurately, generating relevant and customizable financial insights according to each cultural and economic context. Thanks to this iterative process, our model is capable of:
- Interpreting spending patterns and categorizing transactions with greater accuracy in different markets.
- Adapting to local banking systems and financial regulations without losing efficiency.
- Processing data in milliseconds, ensuring real-time analysis regardless of location.
The new era of digital banking experience
At Coinscrap, we are revolutionizing the way knowledge is extracted from unstructured data. Our PoCs use advanced transfer learning techniques to detect complex patterns in text without the need for preprocessing, leveraging multi-layer neural network architectures.
This end-to-end system operates autonomously, eliminating manual intervention and enabling financial institutions to discover valuable insights from data that previously seemed chaotic. What’s truly innovative is how we transform the knowledge obtained from users into information clusters that feed into our deep learning model.
This allows us to build user panels with very specific characteristics, further optimizing categorization and data enrichment. Thus, not only is a traditionally manual and costly process automated, but we also lay the groundwork for scalable solutions adaptable to different languages and contexts.
Our commitment to an AI model capable of reusing the same algorithm across multiple languages –such as Spanish, Portuguese, and English– significantly reduces response times and enhances efficiency. This technological leap not only boosts our offering but also opens the door to serving top-tier clients like Santander Bank.
In short, this project is the first step of an ambitious R&D investment that, in the short and medium term, will result in personalized recommendation products, transforming how banks interact with their customers.
A final reflection: Algorithms for more personal banking
Machine learning and transfer learning have not only driven our international expansion but also strengthened our position as leaders in fintech innovation. The ability to quickly and accurately transfer financial solutions to different markets is redefining the global banking landscape, offering more efficient and personalized services for users worldwide.
With COCO constantly evolving, the future of artificial intelligence applied to banking goes beyond a mere promise. Our technology is leading institutions toward an optimized recommendation system that guarantees increased profits while improving customer satisfaction with digital platforms.
About the Autor
Óscar Barba is co-founder and CTO of Coinscrap Finance. He is an expert Scrum Manager with more than 6 years of experience in the collection and semantic analysis of data in the financial sector, classification of bank transactions, deep learning applied to stock market sentiment analysis systems and the measurement of the carbon footprint associated with transactional data.
With extensive experience in the banking and insurance sector, Óscar is finishing his PhD in Information Technology right now. He is an Engineer and Master in Computer Engineering from the University of Vigo and Master in Electronic Commerce from the University of Salamanca. In addition, Scrum Manager and Project Management Certificate from the CNTG, SOA Architecture and Web Services Certificate from the University of Salamanca. He recently obtained the ITIL Fundamentals certification, a recognition of good practices in IT service management.
