Recommender systems on the Internet

Mnoge internetske usluge kao što su Amazon, Deezer i Netflix koriste sustave za preporuku (eng. recommender systems) kako bi korisnicima pružili preporučene personalizirane sadržaje. Ovakvi sustavi mogu postati i sastavni dio aplikacija. Cilj ovog rada je bio predstaviti sustave za preporuku u kontekstu u kojem se koriste na internetu, njihove vrste i primjene te utvrditi stavove mladih ljudi u dobi od 16-30 godina o sustavima za preporuke, ali i saznati koliko su upoznati s ovim sustavima, koriste li određene internetske usluge isključivo zbog preporuka te smatraju li da im je preporučen sadržaj relevantan. U teorijskom dijelu rada prikazani su osnovni principi rada i algoritama sustava za preporuku, predstavljena je povijest razvitka sustava za preporuke, objašnjene su razlike između primijenjenih tipova filtriranja (kontekstno utemeljeni sustavi, kolaborativno filtriranje, filtriranje temeljeno na znanju, demografski i hibridni sustavi) i algoritmi, te predstavljeni načini i metrike evaluacije sustava za preporuku. Navedeni su primjeri korištenja sustava za preporuku na popularnim internetskim uslugama Netflix i Deezer. U praktičnom dijelu prikazani su rezultati provedenog istraživanja kojim su ispitani stavovi i upoznatost mladih sa sustavima za preporuku te njihov utjecaj na korištenje odabranih internetskih usluga. Na kraju su izrađeni modeli kontekstno utemeljenog sustava i kolaborativnog filtriranja u programskom jeziku Python.Numerous internet services such as Amazon, Deezer and Netflix use recommender systems to provide users with recommended personalized content. These systems can become an integral part of application software. The aim of this thesis is to present recommender systems within the framework of the Internet, enumerate recommender types and applications, and to identify opinions about this topic of young people aged 16-30, as well as to find out the extend of familiarity with recommenders, whether they use any Internet service exclusively because of its recommendations and whether they find the recommended content relevant. In the theoretical segment of this paper the basic principles and algorithms of recommender systems are presented, an outline of the historical development of recommender systems is shown, and types of recommender systems and algorithms are differentiated (content-based recommendation, collaborative filtering, knowledge-based, demographic, hybrid systems), and methods and metrics of recommender system evaluation are given. Then, examples of recommender systems used by popular internet services Netflix and Deezer are given. In the research segment of the paper the results of a conducted survey on the attitudes and awareness of young people about recommender systems are presented, and on their impact on the usage of chosen Internet services. Lastly, a content-based and a collaborative filtering-based recommender system were created in the Python programming language