Improving collaborative filtering with fuzzy clustering
| dc.contributor.advisor | Steel, Sarel Johannes | en_ZA |
| dc.contributor.author | Beukman, Erika | en_ZA |
| dc.contributor.other | Stellenbosch University. Faculty of Economic and Management Sciences. Dept. of Statistics and Actuarial Science. | en_ZA |
| dc.date.accessioned | 2021-11-30T09:18:28Z | |
| dc.date.accessioned | 2021-12-22T14:25:22Z | |
| dc.date.available | 2021-11-30T09:18:28Z | |
| dc.date.available | 2021-12-22T14:25:22Z | |
| dc.date.issued | 2021-12 | |
| dc.description | Thesis (MCom)--Stellenbosch University, 2021. | en_ZA |
| dc.description.abstract | ENGLISH SUMMARY : Recommender systems are machine learning algorithms widely used across various industries to predict user preference for sets of items in order to recommend items to the user. Since it narrows down the entire space of items to a list of items that the client might prefer, it can be seen as an information filtering system. The main purpose of this is twofold: firstly, to introduce new items to users that they might not have otherwise come across, thereby increasing user engagement with products and services, and secondly, to improve user experience. The focus in this report is on collaborative filtering, one of the main approaches to the recommender system problem. A broad range of collaborative filtering techniques is available, including the use of factorization machines. This technique is studied in the research. Factorization machines offer several advantages, one of which is the ease with which information outside of the traditional ratings matrix can be included into the filtering system. Output generated from a fuzzy clustering of users is investigated within the context of a movie recommendation scenario. The positive role which these variables can play in a recommender system is clearly illustrated. | en_ZA |
| dc.description.abstract | AFRIKAANSE OPSOMMING : Aanbevelingstelsels is masjienleer algoritmes wat dikwels in verskillende industrieë gebruik word om die voorkeure van verbruikers vir verskillende versamelings items te voorspel. Hierdie voorspellings word dan gebruik as grondslag vir die aanbeveling van items aan die gebruikers. Aangesien só ’n aanbevelingstelsel die volledige lys items reduseer tot ’n veel kleiner lys van aanbevole items vir ’n gebruiker, kan dit as ’n inligting filtrering stelsel beskou word. Die vernaamste doelwit hiermee is tweërlei: eerstens, om gebruikers bloot te stel aan nuwe items waarmee hulle nie andersins te doen sou kry nie en om sodoende die betrokkenheid van gebruikers by produkte en dienste te verhoog, en tweedens, om die ervaring van ’n gebruiker van die stelsel te verbeter. Die fokus van hierdie verslag is kollaboratiewe filtrering, een van die belangrikste metodes in aanbevelingstelsels. Daar is ’n wye verskeidenheid metodes wat vir kollaboratiewe filtrering gebruik kan word. Faktoriseringsmasjiene, wat in hierdie navorsing bestudeer word, is een hiervan. Faktoriseringsmasjiene bied verskeie voordele, onder andere ’n maklike manier om inligting bo en behalwe dit wat uit die graderingsmatriks verkry kan word, in die aanbevelingstelsel benut kan word. In die verslag word daar ondersoek ingestel na die manier waarop die resultate van ’n nie-besliste (“fuzzy”) segmentasie van die gebruikers in die aanbevelingstelsel ingesluit kan word. Die positiewe uitwerking hiervan in ’n aanbevelingstelsel word duidelik aangetoon. | af_ZA |
| dc.description.version | Masters | |
| dc.format.extent | xii, 155 pages ; illustrations, includes annexures | |
| dc.identifier.uri | http://hdl.handle.net/10019.1/123856 | |
| dc.language.iso | en_ZA | en_ZA |
| dc.publisher | Stellenbosch : Stellenbosch University | |
| dc.rights.holder | Stellenbosch University | |
| dc.subject | Recommender systems (Information filtering) | en_ZA |
| dc.subject | Supervised learning (Machine learning) | en_ZA |
| dc.subject | Groupware (Computer software) | en_ZA |
| dc.subject | Fuzzy decision making | en_ZA |
| dc.subject | Fuzzy clustering | en_ZA |
| dc.subject | UCTD | |
| dc.title | Improving collaborative filtering with fuzzy clustering | en_ZA |
| dc.type | Thesis | en_ZA |