發展商圈行程之個人化推薦系統

Abstract

休閒觀光旅遊與消費型態的轉變，使得近年來商圈觀光旅遊逐漸蓬勃發展，商圈除了滿足在地居民生活機能所需之外，已朝向觀光消費市場轉型。隨著商圈規模的擴大，商圈內的店家與景點琳瑯滿目，人們接收到的資訊十分繁雜，在旅遊、消費的過程中，若想要有趟精緻、完善的商圈旅遊行程，消費者往往需要自行整合大量資訊後再決定行程，造成人們時間成本、精神成本上的額外負擔增加。在這強調個人化的時代，若能考慮不同使用者其不同的偏好，並對其做出一對一的行程推薦服務，便能讓使用者在商圈觀光旅遊時，有效的降低消費者的觀光成本，使其效用最大化，增加消費者的觀光旅遊動機及消費，連動開發商圈龐大的商機，促進地方經濟，帶動地方發展。 本研究旨在建構一套可於行動電子裝置上操作之商圈行程之個人化推薦系統，依據用戶本身的偏好做出個人化的商圈行程推薦，並且歸納商圈行程應提供給使用者之資訊，在系統建構完畢後，引進信義商圈範例進行系統實作，以及進行問卷調查以供修改系統架構之意見參考。本研究之研究對象為系統已有歷史資料之用戶，且使用平台設為行動式電子裝置。本系統主要分為四大模組，依序為屬性偏好分析、協同過濾、情境及輸入資料篩選、行程規劃。首先利用四種資料來分析使用者的偏好，並轉化成使用者對於POI屬性之偏好，接著透過協同過濾找出使用者喜愛之POI，並利用情境資料及使用者輸入之資料進行再一次的篩選，最後透過行程規劃演算法產生行程，並呈獻給使用者。此推薦系統結合文字探勘技術、協同過濾(Collaborative Filtering)與內容式過濾(Content-based Filtering)方法，並解決了一般推薦系統常遇到的評價稀疏性問題，且利用屬性偏好作為相似度計算依據，使得POI數目增加時，計算時間並不會跟隨著拉長，並且利用文字探勘技術，分析文字評論，將其轉化為偏好值，另外，本系統也重視使用者所處情境，利用情境資料及輸入資料使推薦之商圈行程更符合使用者需求及偏好!

Recently, the change of tourism and consumption type make the tourism in business district vigorous. Besides satisfying the necessary of residents, business district has transformed toward the tourism aspect. With the expand of business district scale, the companies and the shops are getting more and more various, so the information consumers receive are so complicated. If the consumers would like a fine and complete tour, they often have to aggregate a large amount of information to decide their tour that makes their time and spiritual cost increase. However, in the time people emphasize personalization, if we can consider the preference for different consumers and provide a one-to-one tour recommendation service, consumers can have less cost for tourism and maximize their utility to increase their motivation for tourism and consumption when they visit the business district. And indirectly, it can develop the business opportunities to promote the economy and development of the area of business district. The purpose of this research is developing a personalized recommender system for tours in business district, demonstrating it by the data of Xinyi Business District, and proceduring a questionnaire to collect the advice of system improvement. The research object is users who the system has his or her historical data, and the platform of this system is mobile device. This system mainly has four modules, which is sequentially analysis of attribute preference, collaborative filtering, selection with contextual data and data imported by users and tour planning finally. At first, the system will analyze attribute preference of users through four kinds of data and find out the preference value of POIs for the specific user collaborative filtering. Then it uses the contextual data and data imported by the specific user to filter and select POI he or she likes. After the POI list are completed, the system will plan the tour by our algorithm and present it to users. The system combines text mining techniques, collaborative filtering, content-based filtering and solve the rating sparsity problem which normal recommender systems meet. It calculates the similarity by attribute preference of users so that its calculating time won’t rise severely when the number of POIs increases. It also use text mining to analyze text comments to transfer into preference value. Besides, the system emphasize the context of users by exploiting contextual data and data imported by users so that the tour this system recommend can match the preference and demand of users!