Automated, statistical ways of detection can only seldom unassistedly detect fraud, but they can go a very long way in detecting anomalous behaviour that can be associated with fraud or misuse.
Creating reliable models that raise Suspicious Activity Reports (SARs) is a major leap forward for organizations looking to curb losses from fraud.
No wonder such techniques are pervasive in some industries, such as credit card transactions, but are yet to reach their full potential in most environments.
Analytics Network’s solutions let customers deploy comprehensive and easily maintanable analytical systems aimed at making sure that the anti-fraud efforts that an organization makes are as efficient and effective as possible.
Our approach realizes that fraud can come in every shape and size – so tecniques vary and combine to create a system that will raise awareness on repetition of known patterns as well as unconvering new types and sources of fraud.
Techniques span across:
- Expert rules, which are easily understandable, turning business savvy into automated, realiable engines;
- Predictive supervised models that learn from the past to spotlight similar risk-laden cases;
- Unsupervised models, ranging all the way from simple univariate outlier analysis, to complex clustering and anomaly detection;
- Social network analysis, through the implementation of such models as diffusion and group analysis.
All the information needs then to be managed, so tools such as risk matrices or specifically-built weighting schemes are available to efficiently select cases to review, thus maximizing results while understanding the unavoidable trade-off between human effort in analyzing cases and revealed cases of fraud
Monitoring activities is the final step, spotlighting the efficiency and effectiveness of each single model as well as of the business decisions involved, therefore creating the grounds for a genuine continuous improvement loop.