diplomsko delo
Gašper Slapničar (Author), Zoran Bosnić (Mentor)

Abstract

Priporočilni sistemi so vseprisotna tehnologija na spletu in lahko ključno vplivajo na poslovne rezultate podjetij. V diplomskem delu se soočimo z razvojem produkcijskega priporočilnega sistema za spletno stran, ki ponuja ekološke nastanitve, ki dosegajo določene ekološke standarde (npr. uporaba sončne energije, filtriranje in ponovna uporaba vode, recikliranje odpadkov itd.). Najprej pregledamo tehnologije velikih podatkovnih množic (angl. big data) in ponudnike strojnega učenja v oblaku. Nato izberemo najustreznejšo platformo in jo uporabimo za zbiranje podatkov in razvoj priporočilnega sistema, ki vrača priporočila za uporabnika z uporabo algoritma matrične faktorizacije (angl. Alternating Least Squares, ALS) ter obenem vrača tudi podobne priporočilne objekte z uporabo Jaccardove podobnosti in evklidske razdalje. Na koncu sistem interno ocenimo na že zbranih podatkih z uporabo statistične mere Precision@k. Rezultati evalvacije so pokazali 19% točnost napovedi, kar je bistveno boljše od naključnega priporočanja, ki doseže 1% točnost. Predlagamo tudi možno implementacijo na spletni strani z namenom izboljšanja poslovnih rezultatov.

Keywords

priporočilni sistem;vzporedno računanje;strojno učenje;velike podatkovne množice;matrična faktorizacija;računalništvo;računalništvo in informatika;univerzitetni študij;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [G. Slapničar]
UDC: 004.85:338.488.2(043.2)
COBISS: 1536565699 Link will open in a new window
Views: 1755
Downloads: 479
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Recommending accommodations using machine learning provider in a cloud
Secondary abstract: Recommender systems are present almost everywhere on the web and can be the key to potentially improved business results. In this thesis we develop a production-ready recommender system for a website that offers eco-sustainable accommodations, that meet certain requirements (e.g. usage of solar energy, water filtering and reuse, waste recycling etc.). First we examine crucial big data technologies and some of the cloud-based machine learning platforms. We proceed to choose the best platform and use it to collect data and develop a recommender system, which returns predictions for a user, based on a matrix factorization algorithm (Alternating Least Squares, ALS). It also returns similar items based on Jaccard similarity and euclidian distance. We conclude with system evaluation by using Precision@k statistical measure. The evaluation results have shown 19% precision accuracy, which greatly exceeds the results of random recommendation that achieves 1% precision accuracy. We also propose a potential website implementation with the intention of improving business results.
Secondary keywords: recommender system;parallel computing;machine learning;big data;matrix factorization;computer science;computer and information science;diploma;
File type: application/pdf
Type (COBISS): Bachelor thesis/paper
Study programme: 1000468
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 60 str.
ID: 8966325