Real-time adaptive morphing website modeled per user and optimized across users

Abstract

Morphing is a powerful tool for providing users with information in a format that benefits them most. It has been shown to increase trust and sales. This thesis describes the implementation of a modular website that morphs based on the click stream of each individual user and learns how to pick the optimal morph based on aggregate user results. The main components are the website controller, the Bayesian Inference Engine, and the Gittins' Optimization Engine. The website controller acts as the interface between the user input and the mathematical modeling of the user's cognitive styles. It uses the Bayesian Engine to update the model and the Gittins' Engine to select the best morph in order to modify the website view. The project was run in survey format to test the effectiveness of morphing for the Suruga Card Loan advice site as well as to test performance and feasibility of real-time morphing and optimization.