AI Module for Customer Preferences System: Product-Level Recommendation Classification with LP-based UTADIS and Machine Learning Benchmarks

Aigerim Kabyl

AI Module for Customer Preferences System: Product-Level Recommendation Classification with LP-based UTADIS and Machine Learning Benchmarks.

Rel. Guido Perboli. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Gestionale (Engineering And Management), 2026

Preview	PDF (Tesi_di_laurea) - Tesi Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (1MB) \| Preview
	Archive (ZIP) (Documenti_allegati) - Altro Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (322kB)

Abstract

This thesis is dedicated to the development and evaluation of the AI module for the customer preferences system. The main task of the work is to classify products at the product level as recommended or non-recommended based on customer reviews and product metadata. The Amazon Reviews dataset for the Electronics category is used for the experiment. This dataset is suitable for this task because it contains both customer reviews and ratings, as well as product metadata. Ratings are converted to customer preference signals, after which the data is aggregated at the product level and combined with metadata. The input criteria are review-based indicators, metadata completeness features, price, brand, and manufacturer information.

The main reference method in the work is LP-based UTADIS, a method from the field of Multicriteria Decision Aid (MCDA) that constructs an additive utility function and a classification threshold