- 學位論文
情境資訊對智慧型裝置上餐廳推薦系統的影響分析
Impact Analysis of Contextual Information in a Mobile Restaurant Recommender System
摘要
民以食為天，吃飯是日常生活中最重要的事之一。除了填飽肚子之外，如何吃、吃甚麼在現今這個社會是個多選項的問題，街上林立的餐廳象徵著生活的便利，同時也意味著人們需要花更多的時間思考到哪用餐。本論文設計了一個可以在智慧型手機上運行的餐廳分享與推薦系統，透過累積眾人過往的經驗(Collaborative Filtering)來幫助使用者快速的尋找到適合的餐廳，並藉由收集到的情境資訊來增進系統效率。 本論文在台灣大學附近進行一個為期兩個月共十七名受測者與六十八間餐廳的用餐資料蒐集，分析的結果顯示出，若一次推薦六個餐廳，在未考慮情境資訊下，推薦系統對使用者真實選用餐廳的命中率僅為26%，而參考情境資訊則能使命中率提升到66%，這驗證了應用情境資訊增進餐廳推薦系統效能的可能性。
並列摘要
There are many restaurants conveniently located in many neighborhoods, as eating out has become an integral part of the modern life. However, deciding what/where to eat can be a challenge for many people. This research develops a restaurant recommendation system on a mobile phone. Collaborative filtering technique is used to make restaurant recommendations to the users. In addition, contextual information sensed by the smart phone is taken into to consideration to improve the quality of recommendations. In this experiment, data for restaurants around the National Taiwan University campus were collected for two months. Facts and reviews of sixty-eight restaurants were contributed by seventeen test subjects. The results showed significant improvement in recommendation quality when contextual information is applied. For a recommendation list of length six,the accuracy grows from 26% without contextual information to 66% with contextual information.