Computer-implemented method and system for testing a model

A computer system comprising:

• a memory configured to store an accurate model indicating a parametric relationship between a plurality of channels of multivariate data and a real dataset of observed multivariate data comprising the plurality of channels;
• a control dataset generator configured to generate a control dataset of multivariate data comprising the plurality of channels, the control dataset being generated based on the accurate model, wherein the control dataset of the multivariate data comprises transaction multivariate data about each transaction of a plurality of transactions performed with a plurality of points of sales (POSs) that each include a processor and a memory, the transaction multivariate data comprising the plurality of channels including a channel that indicates a state of a respective POS of the plurality of POSs handling a transaction of the plurality of transactions, wherein the accurate model indicates a parametric relationship between the state of the respective POS and one or more other channels of the plurality of channels;
• a p-value calculator configured to, for each of a first plurality of sample subsets from the real dataset and a second plurality of sample subsets from the control dataset, calculate a p-value;
• a model accurateness determiner configured to
• generate a first distribution of the p-values of the first plurality of the sample subsets from the real dataset,
• generate a second distribution of the p-values of the second plurality of the sample subsets from the control dataset,
• determine whether (1) a distribution difference characteristic of the first distribution and the second distribution falls within a first predetermined significance range,
• determine whether (2) the p-values for the first plurality of the sample subsets of the real dataset fall within a second predetermined significance range,
• determine that the accurate model is accurate if condition (1) and condition (2) are met, and by iteratively generating and testing a plurality of models;
• a prediction generator configured to if the accurate model is determined to be accurate, predict, with the accurate model, a transaction behaviour for a first POS of the plurality of POSs based on a state of the first POS, wherein the transaction behaviour includes a network resource requirement including one or more of a bandwidth or processing capacity of the first POS;
• and
• wherein a computer-implemented payment authorisation network is configured to allocate the network resource requirement based on the accurate model predicting the network resource requirement, wherein the control dataset generator, the p-value calculator, the model accurateness determiner and the prediction generator are implemented by one or more processors.

A non-transitory computer-readable storage medium comprising instructions which, when executed by a computer system comprising at least one processor, cause the computer system to:

• obtain a real dataset of observed multivariate data comprising a plurality of channels of multivariate data;
• generate a control dataset of multivariate data comprising the plurality of channels, the control dataset being generated based on an accurate model indicating a parametric relationship between the plurality of channels, wherein the control dataset of the multivariate data comprises transaction multivariate data about each transaction of a plurality of transactions performed with a plurality of points of sales (POSs) that each include a processor and a memory, the transaction multivariate data comprising the plurality of channels including a channel that indicates a state of a respective POS of the plurality of POSs handling a transaction of the plurality of transactions, wherein the accurate model indicates a parametric relationship between the state of the respective POS and one or more other channels of the plurality of channels;
• for each of a first plurality of sample subsets from the real dataset and a second plurality of sample subsets from the control dataset, calculate a p-value indicating a parametric relationship between the plurality of channels;
• generating a first distribution of the p-values of the first plurality of the sample subsets from the real dataset;
• generating a second distribution of the p-values of the second plurality of the sample subsets from the control dataset;
• determine whether (1) a distribution difference characteristic of the first distribution and the second distribution falls within a first predetermined significance range;
• determine whether (2) the p-values for the first plurality of the sample subsets of the real dataset fall within a second predetermined significance range;
• determining that the accurate model is accurate if condition (1) and condition (2) are met, and by iteratively generating and testing a plurality of models;
• if the model is determined to be accurate, predict, with the model, a transaction behaviour for a first POS of the plurality of POSs based on a state of the first POS, wherein the transaction behaviour includes a network resource requirement including one or more of a bandwidth or processing capacity of the first POS,
• wherein a payment authorisation network is configured to allocate the network resource requirement based on the accurate model predicting the network resource requirement.