Enhancing B2B E-Commerce with a Tailored Product Recommendation System

Abstract

This bachelor’s project report presents the development and implementation of a tailored product recommendation system to enhance the B2B e-commerce platform Parttrap ONE. Leveraging machine learning models, specifically matrix factorization, the system is designed to analyze historical order data from customer websites to generate personalized product recommendations. The project focuses on evaluating and integrating Microsoft’s ML.NET technologies within a .NET environment, and assess its applicability within B2B e-commerce. Key contributions include the creation of a data preprocessing pipeline, the implementation of an autonomous model training and updating mechanism, and the evaluation of model performance using precision at K (P@K) as a primary metric. The prototype system successfully demonstrates significant results in recommendation accuracy and scalability, although certain limitations such as incomplete GDPR/CCPA compliance and the need for client feedback were identified. Future work will address these limitations and explore additional features to further enhance the system’s capabilities and integration flexibility.