Machine learning-based determination of limits on merchant use of a third party payments system

A system comprising:

• a data repository storing
• a machine learning model,
• a machine readable vector,
• a hyperparameter of the machine learning model,
• an optimized value for the hyperparameter,
• an adjusted hyperparameter of the machine learning model,
• a risk score comprising an estimated probability of a merchant being unable to satisfy an obligation of using an electronic payments system to process a plurality of transactions with a plurality of customers, and
• input data from a plurality of disparate data sources, wherein the input data describes the merchant and an application by the merchant to use the electronic payments system;
• a business rules engine configured to:
• receive the input data prior to calculation of the risk score;
• receive the risk score after calculation of the risk score;
• and
• limit, based on the risk score, use of the electronic payments system by the merchant;
• a featurization layer configured to:
• perform featurization on the input data to generate the machine readable vector;
• a machine learning layer configured to:
• determine, using Bayesian optimization, the optimized value for the hyperparameter of the machine learning model;
• adjust the hyperparameter of the machine learning model to form the adjusted hyperparameter that matches the optimized value;
• receive the machine readable vector;
• execute the machine learning model using the adjusted hyperparameter and further using the machine readable vector as input;
• and
• predict the risk score as an output of the machine learning model.

A non-transitory computer readable medium comprising computer readable program code, the computer readable program code for causing a computer system to:

• receive, in a business rules engine of a computing system, input data from a plurality of disparate data sources, wherein the input data describes a merchant and an application by the merchant to use an electronic payments system for processing a plurality of transactions between the merchant and a plurality of customers;
• perform featurization on the input data to form a machine readable vector, wherein featurization comprises transforming the input data into the machine readable vector;
• determine, using Bayesian optimization, an optimized value for a hyperparameter of a machine learning model;
• adjust the hyperparameter of the machine learning model to form an adjusted hyperparameter, wherein the adjusted hyperparameter matches the optimized value;
• execute the machine learning model using the adjusted hyperparameter, wherein the machine learning model takes as input the machine readable vector, wherein executing the machine learning model outputs a prediction of a risk score, wherein the machine learning model is trained using training data describing use of the electronic payments system by a plurality of other merchants, and wherein the risk score comprises an estimated probability of the merchant being unable to satisfy an obligation of using the electronic payments system;
• limit, by the business rules engine based on the risk score, use of the electronic payments system by the merchant.