Machine learning-based content predictor

A computer-implemented electronic publishing content selection system for use in relation to online electronic transactions, said electronic publishing content selection system comprising:

• a widget encoded on a webpage displayed on a user computing device during an online electronic transaction, said widget configured to:
• send a widget request, via a communications network, for content to be displayed on said webpage, and
• display content on said webpage based on a received selector response;
• and
• a computer-implemented content selection engine coupled to said communications network, said content selection engine including:
• (i) a selector module coupled to a customer data platform that stores customer information, wherein, on receipt of a widget request, said selector module sources customer information from said customer data platform and sends a predictor request based on said customer information and a set of available content items;
• and
• (ii) a programmed content predictor engine configured, on receipt of said predictor request, to:
• select, from a predefined set of page variants, a page variant for said webpage, the page variant defining a number of content placements, each content placement having a number of content positions;
• build a permutation quality table for said webpage, based on:
• (i) said selected page variant;
• and
• (ii) said set of available content items;
• wherein said each entry in the permutation quality table corresponds to a unique content position-content item combination;
• populate each entry in said permutation quality table with a quality score determined by a quality score machine learning model;
• and
• determine a page score for each permutation of content position-content item combinations, based on said quality scores;
• select a permutation of content position-content item combinations having a highest page score;
• and
• send a predictor response to said selector indicating which content items are to be displayed in the respective content positions of the webpage, based on the selected permutation having the highest page score;
• wherein on receipt of said predictor response, said selector module forwards a selector response to said widget, via said communications network, indicating the content items to be displayed in the respective content positions of the webpage.