The Effects of Moods Induced by Webpage Content on the Effectiveness of Display Ads

網頁內容引發之情感對展示型廣告效果的影響

在圖形理論中，相鄰矩陣為表示網絡的基本資料結構；而在線性代數中，一個矩陣經由特徵分析則可找出其特徵值與對應之特徵向量。因此，透過相鄰矩陣的特徵分析將可協助網絡拓撲結構的解析：若綜觀整體特徵向量，可探索整體網絡的聚合次團體；若微觀特徵向量的個別元素值，則可分別評估各個節點的中心性。由於無向網絡的相鄰矩陣為對稱、可以對角化，目前已有許多特徵分析的研究結果。然而，由於有向網絡的相鄰矩陣不一定對稱、不一定可以對角化，所以其特徵分析的研究結果仍然相當有限。因此，本研究嘗試以相鄰矩陣的特徵分析，探索有向網絡的拓撲結構，包含強連通圖形、單一領結結構、以及遞迴領結結構等，進而瞭解有向網絡的相關特徵性質，據以提出一個有向網絡特徵分析演算法，並實務應用在部落格好友網絡，探索其中的拓撲結構。

In graph theory, the adjacency matrix is the basic data structure to represent a network; in linear algebra, the eigenvalues and corresponding eigenvectors of a matrix can be found out via an eigen analysis. Therefore, the eigen analysis of adjacency matrix is an approach to investigating topological structures of networks. After an eigen analysis of adjacency matrix, by inspecting the eigenvectors from the macro view, we can explore cohesive subgroups of the whole network; and by inspecting the elements of eigenvectors from the micro view, we can evaluate centrality of the individual nodes. In the literature, there have been many eigen analytical results of undirected networks thanks to symmetric and diagonalizable adjacency matrices. However, adjacency matrices of direct networks may be asymmetric and not diagonalizable so that the eigen analytical results of direct networks are rare to date. Therefore, in this paper, we try to apply the eigen analysis of adjacency matrix to explore topological structures of directed networks, including the strongly connected graph, simple bowtie structure, and recursive bowtie structure, to further the understandings of eigen properties of directed networks. As a result, we propose an eigen analysis algorithm for directed networks in theory and apply the algorithm to explore topological structures of blogroll networks in practice.