摘要 |
In an e-commerce risk screening system (900), transaction information is applied to multiple fraud risk models (915, 916, 918, 920) that produce raw scores, which are transformed with sigmoidal transform functions (910) to produce optimized likelihood of fraud risk estimates. Such estimates are combined using fusion proportions, producing a single point risk estimate, which is transformed with a sigmoidal function (910) to produce an optimized single point risk estimate. |