| Summary: | 碩士 === 國立雲林科技大學 === 資訊管理系碩士班 === 90 === With the explosion and the rapidly growing market of the Internet, it is imperative that managers re-think to using technology, especially internet, to deliver services faster, cheaper, and with better quality than their competitors do. The web site provides a communication way that reveals real-time assess data and fruitful information of customers. Therefore, the call for customer with personalized web pages has become loud.
To achieve personalized web pages, this study proposed two recommendation algorithms, web page-oriented and user behavior-oriented; by using the web log files from national museum of history. Meanwhile, the following key steps were conducted: (1) To identify the 143 user sessions and 46 web page items. (2) To compute the similarity between the web page items and identify relations between user access patterns. (3) Using Fuzzy C-means to cluster items that exhibit similar click frequency and similar access behavior of users. (4) To save active user’s session through cookie and compute the matching degree between clusters that derived from step 3. (5) To recommend Top-10 page items to active user based on weighting score. (6) To evaluate the quality of the two recommendation algorithms.
The results showed that user-based recommendation algorithm recall value (94.32%) is more accurate than the page-based (79.84%). It is hopefully that the recommendation algorithms presented in this paper may provide valuable information for administrator to manage the web site more efficiency and for users to assess the interested pages more quickly.
|
|---|
