School of Business publications portal
This portal is no longer updated. Aalto University School of Business Master's Theses are now in the Aaltodoc publication archive (Aalto University institutional repository)
School of Business | Department of Information and Service Economy | Information Systems Science | 2014
Thesis number: 13707
Recommendation systems - technology and business aspects
Author: Stenberg, Markus
Title: Recommendation systems - technology and business aspects
Year: 2014  Language: eng
Department: Department of Information and Service Economy
Academic subject: Information Systems Science
Index terms: tietojärjestelmät; information systems; liiketalous; business economics; mallit; models; tietämyksenhallinta; knowledge management
Pages: 73
Key terms: big data; recommendation; recommendation system; prototype; business model
This thesis describes what recommendation systems are, what they are used for, how to build one, and what the potential business models around them are.

The work starts with a literature review of recommendation system usage, recommendation system algorithms, and how to evaluate them. As empirical part of the thesis, implementation effort of a prototype recommendation system is described. The prototype is then evaluated in terms of scalability and quality of recommendations.

Business part of the thesis discusses business models that are based on focusing around different parts of a typical e-commerce process which employs recommendation systems. The models are examined in two dimensions: customer segment type (business or consumer), and for company of what size they are applicable for.

The main conclusion of this thesis is that recommendation systems seem an important part of business. They also seem not very hard to implement, although selling one seems challenging based on failure to sell the produced prototype. It seems puzzling that there does not seem to be much business or even academic literature along the lines of business models noted within the thesis.

In technical implementation, there are obviously some pitfalls, and as the prototype never reached product quality, they may be perhaps under-emphasized. Some interesting work was produced on how to actually design and implement a highly scalable hybrid recommendation algorithm, with linear computational complexity scaling based on number of users, items and rankings to be processed.
Master's theses are stored at Learning Centre in Otaniemi.