The development of a recommendation system on Turi platform

Veliki spletni trgovci ponujajo na milijone izdelkov. Pregledno iskanje po njih strankam predstavlja velik izziv. Priporočilni sistemi so postali ena izmed rešitev tega problema, saj tak sistem strankam predlaga izdelke, za katere obstaja večja verjetnost, da jih bodo zanimali. V tej diplomski nalogi smo opisali in predstavili priporočilne sisteme, opisali najbolj uporabljane algoritmične pristope k priporočanju izdelkov, vzorce razvoja kompleksnejših priporočilnih sistemov in načine, kako ovrednotimo priporočilni sistem. Najbolj smo se osredotočili na gradnjo priporočilnega sistema s platformo za strojno učenje Turi.Large online merchants offer millions of products. A transparent search products is becoming a great challenge for customers. Recommendation systems have become one of the solutions to this problem, since such a system offers products that are more likely to be interesting for the user. In this Thesis we have described and presented recommendation systems, described the most used algorithmic approaches to product recommendations, patterns of development of more complex recommendation systems, and evaluation of the recommendation systems. Most of all, we have focused on building a recommendation system with the Turi machine learning platform