應用深度學習分析評論與地理區域特徵之興趣點推薦方法

Point-of-Interest Recommendation Based on User Reviews and Geographic Area Features by Using Deep Learning Methods

郭伊謙(YI-CHIEN KUO)

指導教授 ： 賴錦慧

中原大學/商學院/資訊管理學系/碩士(2022年)

https://doi.org/10.6840/cycu202201640

摘要

近年來行動裝置的普及，使用者不論任何時間或任何地點，可在社群網路分享打卡行為。這也造成使用者難以找到所需的資訊。為解決資訊過載的問題，許多研究提出興趣點(Point-of- Interest; POI)推薦系統或方法，預測使用者未來可能感興趣的POI。有些研究是分析使用者到訪過的POI標籤、類別、地理位置、打卡時間等因素來做推薦。然而，這些特徵較特定，無法精確表示使用者與POI的特徵。因此，本研究提出一個以評論和地理區域特徵為基礎之興趣點推薦方法(PRRG)，預測使用者可能感興趣的POI。研究架構分為二大部份: (1)評論文字分析與(2)POI區域分析。藉由評論文字分析，可以萃取評論文字中的主題、情感以及語意特徵，用以代表使用者喜好特徵與POI特徵。而POI區域分析則先將POI分成區域，再根據使用者的移動軌跡來計算POI區域權重。最後透過加權矩陣分解法來預測POI之評分。本研究所提的評分預測方法可以萃取多種特徵代表使用者喜好與POI特徵，同時也根據使用者移動軌跡與POI區域對使用者之重要性，來提升POI推薦的準確率。實驗結果顯示本研究方法的表現優於其他方法，並且能有效提升推薦效能。

關鍵字

POI推薦 ； 主題模型 ； 卷積神經網路 ； 閘控遞迴單元 ； 移動軌跡 ； 加權矩陣分解

並列摘要

With the popularity of mobile devices in recent years, users can share their check-in behavior on social networks at any time or anywhere. This also makes it difficult for users to find the information they need. To solve the problem of information overload, many studies have proposed Point-of-Interest (POI) recommendation systems to predict POIs that users may be interested in the future; some of which analyze the POI tags, categories, geographic locations, and check-in times that users have visited to make recommendations. However, these features are too specific to precisely represent the characteristics of users and POIs. Therefore, this study proposes a POI recommendation method based on user reviews and geographic area features (PRRG) to predict the POIs that users may be interested in. The research framework is divided into two main parts: (1) review analysis and (2) POI area analysis. The review analysis extracts the topic, sentiment, and semantic features from user reviews to represent user preferences and POI features. The POI area analysis first divides the POI into areas, and then calculates the POI area weights according to the users’ movement trajectory. Finally, the weighted matrix factorization method is used to predict the ratings of POIs. The proposed method can extract various features to represent user preferences and POI features, and analyze the importance of POI area to users based on their movement patterns to enhance the recommendation accuracy. The experimental results show that the proposed method outperforms other methods and effectively improves the recommendation performance.

並列關鍵字

POI Recommendation ； Topic Model ； Convolution Neural Network ； Gated Recurrent Unit ； Movement Patterns ； Weight-Matrix Factorization

參考文獻

Agrawal, S., Roy, D., & Mitra, M. (2021). Tag Embedding Based Personalized Point of Interest Recommendation System. Information Processing & Management, 58(6), 102690.

Albawi, S., Mohammed, T. A., & Al-Zawi, S. (2017). Understanding of a Convolutional Neural Network. International Conference on Engineering and Technology (ICET), Antalya, Turkey. 1-6.

Alorf, A. (2017). K-Means, Mean Shift, and Slic Clustering Algorithms: A Comparison of Performance in Color-Based Skin Segmentation [Master's Thesis, Doctoral Dissertation, University of Pittsburgh]. Electronic Theses and Dissertations.

Bahdanau, D., Cho, K., & Bengio, Y. (2015). Neural Machine Translation by Jointly Learning to Align and Translate. arXiv preprint arXiv: 1409.0473.

Baral, R., Zhu, X., Iyengar, S., & Li, T. (2018). Reel: Review Aware Explanation of Location Recommendation. Proceedings of the 26th Conference on User Modeling, Adaptation and Personalization, 23-32.