InstanceRank: A Framework for Predicting Ranksfor Instances and Its Application to Customers'' Webpage Click Log Mining

• Main Authors: Ching-Hsiang Wen, 温景翔
• Other Authors: 范耀中
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/3335tk

碩士 === 國立中興大學 === 資訊科學與工程學系 === 106 === In this thesis, we discuss the problem of ranking instances in a set. The goal is to associate each instance with a rank, which is an integer from 1 to k, where k is the number of instances in the given set. To address this problem, we propose a framework called InstanceRank, which predicts each instance a rank based on the partial pair-wise instance comparison results. The idea is to model the pair-wise instance comparison results as a network model, where a node is an instance, and an edge between two nodes is the comparison result, and then apply stochastic process to learn the ranks of instances. We describe two sets of experiments, with eSports ranking data and with the e-commerce shopping-cart viewing/buying logs, to show the effectiveness of the proposed framework.