Model Overview

Matchmeter is a machine learning model trained to distinguish between two classes of name strings: one provided by the merchant and the other from the end user’s bank record.

Class ‘1’ indicates that the two strings correspond to the same person, while class ‘0’ signifies that they do not match. Additionally, Matchmeter returns a metric on a scale from 0 to 100, expressing the degree of similarity between the strings, where 0 denotes completely different strings and 100 indicates a perfect match.

Model Performance

Recall and precision are metrics used for the evaluation of a machine-learning classifier.

  • Recall in simple terms answers the question: Out of all the actual positive (negative) cases, how many did the model successfully detect?

Therefore, recall indicates how effectively the model detects all instances of a given class.

  • Precision in simple terms answers the question: “Out of all the cases that the model predicted as positive (negative), how many were actually positive (negative)?

Therefore, precision indicates how accurate the model’s predictions are for a given class.

Our model has been optimised by default to maximise the recall metric for class ‘0’ (no match) We assume that the primary goal of the model is to detect all cases of incorrect registration or attempted fraud

Matching Example

The effectiveness of Matchmeter has been tested using a realistic dataset that preserves the distribution of bank popularity. However, it’s important to note that the results may vary depending on factors such as i.e. the quality of the shopper’s name provided.

Shopper name                           Account holder name           Score   
Amelia Grace Montgomery Amelia Grace Montgomery 1.0
Amelia Grace Montgomery Montgomery A G 0.7634
Amelia Grace Montgomery Montgomery A 0.75
Amelia Grace Montgomery Montgomer A 0.6216
Amelia Grace Montgomery Montgom A 0.2576
Amelia Grace Montgomery Montgomery 0.2961
Amelia Grace Montgomery M Amelia Grace 0.3544

Matchmeter Enablement

To enable and use Matchmeter you need to ensure the following: 

  • you requested your Implementation Manager or Account Manager the enablement of this Product and completed all the commercial onboarding steps 
  • you authenticate first before you can request a payment
  • you integrate the relevant endpoints