Machine learning techniques have evolved remarkably over the last three decades. At Neural Technologies we’ve been part of that transformation from pioneering theory to delivering effective fraud management, revenue protection, and data integration solutions across a wide range of industries and applications.
The machine learning process offers a sophisticated and differentiated approach that far outstrips the potential of rigid, rules-based decision making for businesses. At Neural Technologies that means leveraging the unique benefits of an ActivML system that can automate processes, guaranteeing quicker and better results for our customers.
ActivML offers beneficial opportunities across the business landscape, from machine learning in telecommunications through to applications in traditional industries such as energy and utilities. Neural Technologies’ ActivML deployment includes three core differentiators that set it apart from more rigid legacy rules-based approaches.
- Self-learning. ActivML provides automatically optimized identification for representative profiles as part of structured analytical profiling. Explained structural information is provided with descriptive differentiating factors to provide data understanding users to analyze why and how a profile is rated. It incorporates a confidence-based approach which maps new data to structural profiles, providing simple change recognition with clear reasoning.
- Unknown detection. ActivML provides unconstrained anomaly detection that can quickly identify emerging threats. Unidentified events are detected and analyzed with confidence and clear reasoning.
- Risk prediction. Structured classification prediction provides clear threat ratings, with separate records classified into multiple target and non-target predictions for rapid analysis. User-friendly analytics provide clear explanation of why outcome classifications are given.
ActivML offers an evolving solution
Like any ‘true’ machine learning applications, ActivML is a system designed to grow and evolve. This means it not only detects preprogrammed threats, but adapts to protect against the ‘unknown unknown’. That’s particularly important in a high-volume data landscape where fraudsters are constantly attempting new avenues of attack.
A training dataset is used as an initial training tool for ActivML, utilizing labels in the case of supervised learning. Automated machine learning is then applied to all data processed by the system, providing clearly explainable outputs such as structural profiles of data, explainable features of the data, classification and clear anomaly detection. This process not only provides robust protection against fraud, but provides clearly explainable analysis to customers.
Since the ActivML model is constantly learning and evolving, the automated machine learning generates a model that learns from its own analysis, meaning classified and anomalous data is then incorporated in a feedback loop that informs future data analysis. That powers a system which is constantly adjusting and re-tuning to offer ongoing improvement.
Implementing the ActivML model for telecommunications organization TurkCell improved fraud detection by 12%, reducing time between analysis and intervention by 27%, and improving collection amounts by 89%.
ActivML implementation was customized for Turkcell to ensure clear insight for analysts, making a user-friendly solution that helped inform decision making in real time. That meant delivering on four simple steps.
- Data and feature selection
- Data gathering
- Training and modelling
- Evaluation and tuning
Neural Technologies has deployed ActivML capabilities for a wide range of industries and applications. That includes active use cases tackling subscription fraud, sim box or bypass fraud, international revenue sharing fraud, and dealer fraud, amongst others.
ActivML capabilities also unlock substantial future opportunities for customers. It offers a rich source of data collected across an organization, which can be leveraged for a wide range of purposes. That includes customer information for more personalized offers, provision, customer relationship management, billing, network services, sales, and much more.
This machine learning solution can also provide insight into overall network usage, which can be a powerful tool for understanding the potential of emerging technologies such as 5G. Machine learning in telecom companies offers a particularly valuable opportunity in this regard.
More broadly, ActivML offers dynamic billing, pricing, e-money, and discount opportunities that can help telecom operators and others address changing customer needs. That means building beyond simply fraud, revenue assurance, and credit, to provide true integrated data opportunities in digital transformation and customer experience that can help operators stay competitive in the modern business landscape.
Want to power up your business with data? Speak to the team at Neural Technologies today.