建構使用者導向的個人化數位電視節目推薦模型

Abstract

數位電視的技術壓縮了訊號傳送的方式，能夠容納多於目前好幾倍的節目頻道，大量的媒體內容讓使用者擁有更多元化的選擇，另一方面也帶來資訊過量的議題。個人化的電視節目推薦系統能夠記錄使用者的偏好資料，以資料探勘的方式篩選並推薦一些符合使用者興趣的節目，幫助使用者解決選擇過多的問題。然而過去的一些個人化推薦系統往往只關注在使用者的收視行為層面上，而忽略了使用者本身的特性。本研究提出一個以使用者為中心的MEDIA 個人化節目推薦模型，除了根據使用者的收視經驗與人口資料之外，還加入了使用者的生活型態以及觀看時的心情來預測使用者的觀看需求，並提供節目的建議，使得數位電視的使用能夠簡易化、個人化。經由類神經網路的分析結果，證實本研究所提出的推薦預測模型較以往的方式更具準確性，可供未來電視推薦系統設計時的參考。

The technology of the DTV has compressed the way of the signal transmissions, so that DTV can contain much more times of channels than before. A mass of media contents allow people to have more kinds of choices, on the other hand, it also comes along with the information overload issue. A personal DTV recommendation system can filter and recommend DTV programs which accord with user's interest by using some data mining methods. It can help users to deal with the problems of too many choices. However, some recommendation systems just focused on the level of user's behavior of DTV watching in the past, and they ignored user's own characteristic. In this research, we propose a user-center based recommendation model – MEDIA model, which considers not only the user's watching experience and demography, but also their lifestyle and the moods when they watching DTV. We use the system to predict user’s need of watching DTV and provide some suitable program suggestions for the users. It will make DTV more friendly and personl to users. Via the analysis of neural network, the results show the recommendation model of this research has more accuracy and correctness than previous models, and it can further designs on the DTV recommendation systems in the future.