A Study of Using Collaborative Filtering for Web-Sites Recommendation
碩士 === 國立臺北科技大學 === 商業自動化與管理研究所 === 89 === As the rapid growth number of WWW websites, it is more difficult to find our interested websites with high quality than before. In general, we search proper websites through search engine service. But there are three serious problems affect search engine se...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2001
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/91009293608369855299 |
| Summary: | 碩士 === 國立臺北科技大學 === 商業自動化與管理研究所 === 89 === As the rapid growth number of WWW websites, it is more difficult to find our interested websites with high quality than before. In general, we search proper websites through search engine service. But there are three serious problems affect search engine service: (1) We must know exactly what key word will lead us to the right direction. (2) A search will sometimes return thousands of “successes.” (3) None of the popular search engines indexes over 50 percent of the Web content, and less than 1 percent of Web pages show up in all of the popular engines. Thus, this study proposes using collaborative filtering to automate the process of “word of mouth” among WWW users. There are only a few researches about collaborative filtering. And most of them face five problems: (1) explicit rating; (2) first-user problem; (3) lack of privacy; (4) rating sparsity; (5) collect object before recommendation. We propose a new mechanism of using collaborative filtering for web-sites recommendation. This mechanism can automatic produce rating value about web sites and avoid traditional problems of collaborative filtering. Finally, we observe “effective recommendation number” and “effective recommendation rate” from the simulate experiments and found that using implicit rating can get better results than explicit rating.
|
|---|