Internetbokhandelns rekommendationssystem : en undersökning av Amazon.coms Similar Items.
| Autor: | Arvidsson, Sofia, Tolstoy, Theodor |
|---|---|
| Jazyk: | švédština |
| Rok vydání: | 2005 |
| Předmět: | |
| Druh dokumentu: | Text |
| Popis: | The aim of this thesis is to examine one part of Amazon.coms recommender systems: Similar Items. The purpose is also to show the difficulties with subject access, i. e. to find a similar book to the one you just read; problems with traditional systems versus the usefulness with recommender systems like Amazon.coms Similar Items. Similar Items presents a set, a cluster of related items to a given item based on customers co-purchase. The questions examined are: How connected are the created clusters of related similar items? How common is it that the author of the seed origin book is also the author of the books in the cluster? What kind of other connections are there in the cluster? How common is it that the similar items in the cluster have the same Browse Node category as its seed? In a more qualitative study of 10 books and their similar items; how similar are they? The study consists of a quantitative study of similarities between items and their similar items in 4 iterations and a qualitative study of 10 books. The data collection was extracted from two categories: African American and Russian. The results for the quantitative part show that the number of similar items that are related to the seed is decreasing after every iteration. At the first iteration the clusters are closely connected. The qualitative results show that the clusters have at least one or two common similarities to their seed or to each other. This concludes that Similar Items as a function could be an appropriate complement to a library online catalogue. With more possibilities of accessing fiction, users should have more success in searching for fiction. Uppsatsnivå: D |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|