Hybrid Movie Recommender System

Sustavi preporuka su podvrsta sustava filtriranja informacija koji nastoje predvidjeti korisnikov stav prema određenom predmetu. S obzirom na pretpostavke i podatke dijele se na sustav preporuka zasnovan na suradničkom filtriranju i sustav zasnovan na sadržaju. Hibridni sustavi preporuka koriste dvije ili više vrsta sustava preporuka i njihovom kombinacijom eliminiraju pojedinačne nedostatke i postižu bolje rezultate. Ovaj rad se sastoji od implementacije sustava preporuke filmova zasnovanog na suradničkom filtriranju, sadržaju i hibridnog sustava preporuke, te vrednovanjem svakog od sustava. Provedeno je mjerenje srednje apsolutne pogreške i pogreške srednjih kvadrata na 5 slučajnih uzoraka i pokazano je da hibridni sustav daje najmanju pogrešku za svaki uzorak. Zaključeno je da hibridni sustavi daju bolje rezultate i eliminiraju nedostatke pojedinačnih tehnika preporučivanja.Recommendation systems are a subgroup of information filtering systems that strive to predict the user's attitude towards the item. According to the hypothesis and data they use, they are divided into a system of recommendations based on collaborative filtering and a system based on content. Hybrid recommendation systems combine two or more types of recommendation systems to eliminate individual failures and to achieve better results. This thesis contains an implementation of a movie recommendation system based on collaborative filtering, a system based on content and a hybrid recommendation system, and also an evaluation of all implemented systems. Mean absolute error and mean square error measures were performed on 5 random samples and it was shown that the hybrid system gives the smallest error for each sample. It was concluded that hybrid systems give better results and eliminate the failures of individual techniques