Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： The existing evaluation methods for node importance in complex network mainly focus on un-directed and un-weighed networks, and can’t describe the real-world network completely. For example, most centrality measures only consider a single indicator, ignoring the difference between out-degree and in-degree of the node, and neglecting the importance of weight. Based on the directed-weighted complex network, this paper proposes a center node recognition algorithm, cw-shell decomposition method, which is based on out-degree, in-degree and weight, considering the difference between out-degree and in-degree of the node, and the practical importance of weight in real network. In order to verify the effectiveness of the new index, the weighted-susceptible-infectious-recovered model is used to simulate spreading process on real-world networks. The results show that the cw-shell decomposition method can rank the nodes efficiently, and identify the nodes with higher diffusion ability.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-17 Cooperative journals: 《计算机应用研究》
Abstract： This paper studies the extraction and clustering of product features in online user reviews. It proposed an improved SimRank algorithm to put the affective word-feature pair into the binary network. And the improved SimRank algorithm is used to compute the similarity between the characteristic words. Then the spectral clustering algorithm is adopted to cluster the feature similarity. Extracts feature sets for network products. Taking a computer commentary as an example, the paper extracts affective word-feature pairs, and the experimental results show that the improved algorithm has higher accuracy. The improved feature similarity detection method can be used as an effective method for detecting feature similarity. The experimental results show that using the improved Sinrank similarity to extract the feature words is more accurate.
Hits
Hits
Downloads
Comment