| Summary: | 碩士 === 國立臺北大學 === 企業管理學系 === 92 === According to "the investigation on Taiwan use of network in 2003", the main puzzle in accessing to internet is too much trash information. After the internet was developed, information overload create user''s puzzle instead of information insufficient in the past. Collaborative filtering can provide personalized information and resolve the problem of information overload. Collaborative filtering successfully has applied to the actual website operation and provides the user individual product recommendation.
Another problem for information overload is the capital estimated excessively which creates the data transmission speed excessively to be slow. According to the identical investigation demonstrated "the transmission speed" is the most important factor which user chooses the website. In order to resolve challenge in the internet, this research utilizes neural network in the recommendation mechanism. Based on the result, this study provides the following conclusions:
1.If using explicit ratings in the collaborative filtering, it causes heavy loading on user to unfold their preference.In our empirical, there is a lower significant positive relationship between user''s rating quantity and prediction accuracy.It alleviate heavy loading on user.
2.Applying method in this research on neighbor formation process reduces process time for prediction user''s preference.The saving time is satisfying with user''s requirement in transmission quality and speed.
According to the result in this research, predicting quickly user''s preference can be the base of recommended commodity.
|
|---|
