fieldid E-Thesis & Research สถาบันเทคโนโลยีไทย-ญี่ปุ่น
สืบค้น:

เขตข้อมูล ข้อมูล
บทคัดย่อ
Development of Analysis of the Accuracy and the Diversity in the Recommender System Based on Collaborative Filtering Approach : One of the current challenges for improving recommender systems is to find an optimal way for diversity and accuracy trade-off. This research aims to find that for real-life educational data, how much impact of diversity for accuracy of the system by developing the collaborative filtering recommender system to conduct experiments and analysis. An analysis showed that using MSD similarity as the similarity calculation method and KNN with Means as the algorithm will give the best prediction result for the user-based system. For the item-based system, using cosine similarity as the similarity calculation method and KNN Baseline as the algorithm will give the best prediction result. Diversity of recommended item affects each system differently. For user-based system, having more choices for the recommended subjects can lead to better prediction result due to there is more data that can be used. For the item-based system, having more choices for a recommended item may not lead to better prediction result due to similar recommended item cannot be used much in the item-based system.
ผู้แต่ง
Vivat Thongchotchat
ประเภทสิ่งพิมพ์
วิทยานิพนธ์ (MIT) 2019
เลขหน้า
125
หัวเรื่อง
Information Technology--Thesis
หัวเรื่อง
Thai-Nichi Institute of Technology.Information Technology--Thesis
เอกสารฉบับเต็ม
Download full text