Third party vendor fraud risk assessment using shapley additive explanations for value attribution

A non-transitory computer readable medium comprising computer readable program code, which when executed by a computer processor enables the computer processor to perform a method for managing value attribution in a data system, the method comprising:

• obtaining, by the data system, order data associated with a plurality of data vendors, wherein the order data comprises features about client devices and about a set of orders;
• in response to obtaining the order data:
• applying the features in the order data to a classification model to obtain fraud prediction models;
• performing an explainable mechanism on the fraud classification predictions and the features using a shapley additive explanation (SHAP) to obtain localized SHAP values each corresponding to one of the features;
• applying a set of custom transformations to the localized SHAP values to obtain normalized features;
• applying a value attribution algorithm on the normalized features using vendor information associated with the features and the plurality of data vendors to obtain vendor values each associated with one of the plurality of data vendors;
• performing a vendor remediation on the plurality of data vendors based on the obtained vendor values,
• wherein the vendor remediation comprises increasing or decreasing resources applied to each of the data vendors based on the vendor values, and
• wherein the resources of each of the data vendors comprise memory and network bandwidth applied by the data system for the each of the data vendors, and
• wherein the resources are increased and decreased based on limiting communication with each data vendor on a granular level.

A system, comprising:

• a processor;
• and
• memory including instructions, which when executed by the processor, perform a method comprising:
• obtaining order data associated with a plurality of data vendors, wherein the order data comprises features about client devices and about a set of orders;
• in response to obtaining the order data:
• applying the features in the order data to a classification model to obtain fraud prediction models;
• performing an explainable mechanism on the fraud classification predictions and the features using a shapley additive explanation (SHAP) to obtain localized SHAP values each corresponding to one of the features;
• applying a set of custom transformations to the localized SHAP values to obtain normalized features;
• applying a value attribution algorithm on the normalized features using vendor information associated with the features and the plurality of data vendors to obtain vendor values each associated with one of the plurality of data vendors;
• performing a vendor remediation on the plurality of data vendors based on the obtained vendor values,
• wherein the vendor remediation comprises increasing or decreasing resources applied to each of the data vendors based on the vendor values, and
• wherein the resources of each of the data vendors comprise memory and network bandwidth applied by the data system for the each of the data vendors, and
• wherein the resources are increased and decreased based on limiting communication with each data vendor on a granular level.