A Dynamic Time Weight-based Collaborative Filtering Recommendation System

• Main Authors: Yu-Shiun Huang, 黃昱勳
• Other Authors: Yi-Cheng Chen
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/yd62d2

碩士 === 淡江大學 === 資訊工程學系碩士班 === 105 === Traditional time weighted collaborative filtering systems have a single decay function. But, it is not reasonable that lets the weight decay by only function. We propose a method to solve it. In this paper, we propose a new method improve on time weighted collaborative filtering. We use the principle of human brain memory to give different time segments corresponding to the recession function: instantaneous memory level, short-term memory level, long-term memory level, whenever there is a new rating come in, and its related item cluster will be activated to a new recession level (instantaneous memory level). We set a threshold value. If the number of activations is less than the threshold for a certain period of time, we will give the corresponding penalty, otherwise we will raise his decay function to a higher level (short-term memory level), and so on. Once the long-term memory level is reached, even if the number of activations does not reach the threshold, there will be no penalty, but will let it fall with his decay function. Finally, the weight of the decay is weighted within the Item-based predictive function, which is a post-processing approach.