20.500.12556/RUL-144045 Avtomatski izbor artiklov za personaliziran e-katalog Automatic selection of products for a personalized e-catalogue V diplomski nalogi smo analizirali in implementirali priporočilni sistem za personaliziran elektronski katalog. Opisali smo uporabljena orodja in delovanje naših programov za transformacijo podatkov in izdelavo e-kataloga. Zapisali smo tudi prednosti, ki jih ponujajo e-katalogi pred tradicionalnimi in ugotovitve, ki so jih zapisali raziskovalci za podoben problem. Za izdelavo e-kataloga smo najprej opisali, kje smo pridobili nabor podatkov, kako smo ga spremenili in omejili za naše potrebe. V nadaljevanju smo opisali namen in implementacijo uporabljenih metod ter rešitve težav, ki se pojavijo pri izdelavi e-katalogov. Na koncu smo napovedi metod združili in prikazali uspešnost izdelanega priporočilnega sistema. Rezultati našega personaliziranega e-kataloga prikazujejo informacije o natančnosti, priklicu in F-oceni. Pri tem nas je zanimal predvsem podatek o priklicu, saj nam pove, koliko izdelkov moramo priporočiti, da zadovoljimo vse interese uporabnikov. Poleg rezultatov priporočilnega sistema smo izvedli teste in prikazali učinkovitost vsake metode. In the diploma thesis, we analyzed and implemented a recommendation system for personalized electronic catalogue. We have described the tools used and the operation of our programs for data transformation and e-catalogue creation. We also wrote down the advantages offered by e-catalogues over traditional ones and the findings written by researchers on a similar problem. To create the e-catalogue, we first described where we obtained the data set and how we modified and limited it for our needs. Next, we described the purpose and implementation of the methods used, as well as solutions to problems that arise when creating e-catalogues. At the end, we combined the predictions of the methods and showed the efficiency of the created recommendation system. Our personalized e-catalogue results show precision, recall and F-score information. We were particularly interested in recall data, as it tells us how many products we need to recommend in order to satisfy all user interests. In addition to the results of the recommender system, we implemented tests and demonstrated the effectiveness of each method. priporočilni sistem skupinsko filtriranje gradientno pospeševanje e-katalog napoved izdelkov recommendation system collaborative filtering gradient boosting e-catalogue products prediction true false false Slovenski jezik Angleški jezik Diplomsko delo/naloga 2023-01-27 17:15:00 2023-01-27 17:15:10 2023-01-30 09:33:49 0000-00-00 00:00:00 2023 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 35088 139878915 Macur_Leon_-_Avtomatski_izbor_artiklov_za_personaliziran_e-katalog.pdf Macur_Leon_-_Avtomatski_izbor_artiklov_za_personaliziran_e-katalog.pdf 1 EA8797C111DAE0A78F225E9234D17064 85a01ab4c21c093876299a6b5f1a079c044f22805fad809b4937cac9f6c070be 82124e0b-9e5d-11ed-9c94-00155dcfd717 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=166167 Fakulteta za računalništvo in informatiko 0 0 0