一個藉由改善稀疏矩陣處理方式之協同過濾式推薦

An Improved Method of Processing Sparse Matrix in Collaborative Filtering Recommendation

陳柏良(BO-LIANG CHEN)

指導教授 ： 李麗華

朝陽科技大學/資訊學院/資訊管理系/碩士(2017年)

摘要

現在網際網路科技的發達，網路資料量日益增多，使得人們逐漸無法處理過於龐大的資料，也因此出現了資訊過載(Information Overloading)的議題。當資料量變得龐大時,不只會帶來資訊過載的問題，在資料儲存上也常會出現因使用者數目龐大、資料項目過多而造成稀疏矩陣，例如：擁有成千上百產品或數以千計書籍的網路書城，其使用者交易的書目通常為個位數，因此”顧客”在數以千計的”產品項目”項下，只能針對少數幾項產品有交易資訊，如此就會形成資料儲存時的稀疏性。
過去有許多學者提到稀疏性問題，然而在協同過濾式推薦研究上，則較少有學者針對此一議題提具出有時效性之解決方法。有鑑於此，本研究提出一個藉由改善稀疏矩陣處理方式之協同過濾式推薦，提升處理稀疏矩陣在運算上的時效性。本研究所提之方法較傳統常用的皮爾森相關係數、歐幾里德距離、傑卡德指數，在時效上與準確度上都有較佳的結果。

關鍵字

推薦系統 ； 協同過濾 ； 稀疏性

並列摘要

As the development and usage of Internet keeps growing, the sheer volume of data on the Internet makes it difficult for user to handle. As a result, users must deal with the information overloading problem. Even when the amount of data keeps booming such as hundreds of products, thousands of books, however, the amount of transactions for each product item will still be similar. Therefore, the transaction data among thousands of items and thousands of uses may form the sparse matrix. For collaborative filtering recommendation, there are many researchers proposing great methodologies to handle user-item-matrix dataset, however, not many researchers deal with the performance of sparse matrix problem. This is the reason why this research proposes a novel solution for collaborative filtering recommendation. The proposed method has compared to the traditional Pearson Correlation Coefficient method, Euclidean Distance, Jaccard index and the results show that our proposed method is better in terms of time efficient and precision.

並列關鍵字

Recommender System ； Collaborative Filtering ； Sparsity Matrix

參考文獻

[1] 蔡美慧，“以偏好導向為相似性量測的協同過濾式推薦”，朝陽科技大學資訊管理系碩士論文，2010。

[27] J. B. Schafer, J. Konstan, and J. Riedi (1999), “Recommender systems in e-commerce,” Proceedings of the 1st ACM Conference on Electronic Commerce, pp. 158-166, Denver, Colorado, USA.

[33] Junrui Yang, Cai Yang, Xiaowei Hu, (2016), “A study of hybrid recommendation algorithm based on user,” 2016 8th International Conference on Intelligent Human-Machine Systems and Cybernetics (IHMSC), pp. 261-264, Hangzhou China.

[34] Peng Yi, Chen Yang, Xiaoming Zhou, Chen Li, (2016), ” A movie cold-start recommendation method optimized similarity measure,” 2016 16th International Symposium on Communications and Information Technologies (ISCIT), pp. 231-234, China Qingdao.

[15] V. Kavinkumar, Rahul Reddy Rachamalla, Balasubramanian Rohit, M. Sridhar, Sridharan.K,Dr. D.Venkataraman, (2015), “A hybrid approach for recommendation system with added feedback component,” 2015IEEE, International Conference on Advances in Computing, Communications and Informatics (ICACCI), pp. 745-752, Kerala Kochi, India .