利用情境感知晶片實踐適地性自動日誌系統

Abstract

近幾年來，智慧型行動裝置蓬勃發展，包含上網速度、運算能力越來越快，以及裝置上的嵌入式晶片越來越豐富，這樣的變化給予行動商務平台甚廣的發展空間，適地性服務就是其中最廣受歡迎的應用之一。 在這篇論文中，我們提出了一個全新的整合服務：我們融合了自動化的日誌紀錄服務以及預測使用者未來事件的功能，提供使用者檢視自身過去歷史、系統對於當前所發生事件的回應、以至於未來即將發生的事情預測，從過去到未來的生活全都涵蓋到了。在自動化日誌紀錄的部分，我們利用手機內建的多種嵌入式晶片識別出使用者所在處最有可能的point of interest，並且運用使用者過往的紀錄，推測出使用者在這個地點的行為。而在未來事件預測的部分，本論文藉助於資料探勘的技術，運用分類與關聯關係分析的方法從使用者的過去經驗中，事先找出未來最有可能發生的項目。

Recently, mobile devices have faster network connection and greater computation power. Also, various sensors have been embedded on to the hand-held devices. This trend has leaded the mobile computing field into a new era and numerous applications can be achieved. Within all, location based services are one of the biggest new hit. In this thesis, we demonstrated a new integrated service called MobiFairy, which we combined the automatic journal service with a personal future event prediction mechanism. With this application, we provided the user the ability to browse through his or her past history, display recommendation according to current events, and predict the user’s future behavior. In short, we’ve covered the whole timeline. In the part where the application automatically does the journaling, we take advantage from the mobile embedded sensors. By monitoring various sensors, we came up with a sensor recognition mechanism which can recognize point of interest of where the user has visited. By matching past history, it is possible for the system to find the user’s most possible event occurred at this current location. In the part of future prediction, this thesis applied data mining techniques including classification and association rules analysis. By using these two methods, the system could learn from the past event patterns of its user, and make the most possibility estimation toward the near future in advance.