Preliminary Conference Agenda

Overview and details of the sessions of this conference. Please select a date or room to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

This agenda is preliminary and subject to change.

Session Overview
Chinese Papers 7: Data Mining & Analysis
Saturday, 25/Mar/2017:
10:30am - 12:00pm

Location: Hanyang Hall
Location: Third Floor Capacity: 60 Size: 73㎡

Show help for 'Increase or decrease the abstract text size'

Predicting missing links in co-authorship networks: a clustering perspective

Bin Zhang1, Yating Li2

1Center of Traditional Chinese Cultural Studies, Wuhan University, Wuhan, China; 2Center for the Studies of Information Resources, Wuhan University, Wuhan, China

This paper selected seven disciplines data from CSSCI database, and used them to construct co-authorship networks. The giant components were extracted for experiment of link prediction, and the result of the experiment would be compared with the change of the network structure with correlation analysis. The research showed that the nearest neighbor’s local similarity indices would be effective when the clustering coefficient is high. With the increase of the clustering coefficient, SRW is better in sparse network, and AA and RA have the greater performance in dense network.

Latent Topic Model based Review Opinion Clustering and Measurement

Hui Nie

Sun Yat-sen University, China, People's Republic of

With the rapid growth of internet, a wealth of user-generated product reviews has been spread to the web. However,product reviews posted at commercial websites vary greatly in quality. In the paper, we attempt to solve the problem of how to extract and integrate the useful content from the massive product reviews, in order to make the most of online reviews useful for commercial application. For the purpose, great attention has been focus on the review summation for Chinese language. The two crucial problems for the review summation are the user-opinion based product aspect clustering and the user-opinion measurement. Opinion_LDA, a Latent Aspect Model based on feature words sequence, has been built to aggregate users’ opinion automatically. Meanwhile, a polarity lexicon and modified relations between terms have been employed for measuring the user opinion. The proposed model and method have been tested in a real tablet compute -specific reviews dataset. The result indicates the effectiveness and feasibility of the method for our review summarization task.

Phishing websites identification with URL abnormal features and online evaluation data

Zhongyi Hu, Chaoqun Wang

Wuhan University, China, People's Republic of

As the e-commerce and e-finance expand, phishing has attracted broad attention in the areas of Internet security. In this study, by applying eight machine learning techniques, the performance of the online evaluation data of web domains in identifying phishing websites are compared with that of URL abnormal features. The potential of improvement of performance is also investigated by combining the aforementioned features. Results show that the evaluation data of domains has better performance in identifying phishing websites than abnormal features of URL. In addition, combination of different kinds of features is effective in improving the identification performance.

The development of key technology for the data extraction in scientific evaluation based on multi-objective optimization programming ideas

Junping Qiu, Lei Han

Wuhan University, China, People's Republic of

The index system in scientific evaluation is huge and complicated, and the data processing has to been firstly finished in the scientific evaluation study. Therefore, a considerable amount of time is spent for dealing the complex data. In this paper, an idea of free programming was proposed to establish the model of data extraction using recursion-tree method on the base of divide-and-conquer strategy in computer algorithm theory. We believed that the compilation and process data could be completed by a more simply and scientific method through using this model, so that the efficiency of scientific evaluation could be greatly enhanced.

Multi-scale and Cross-scale Analysis on Collaboration Network

Ping Liu, Yanan Li

Wuhan University, China, People's Republic of

Scientific collaboration networks have been widely investigated due to the increasingly importance of globalization and international collaborations. Most studies of collaboration networks put an emphasis on the structure of the network, that is, only relationship between nodes is taken into considerations. The information of node attributes is ignored. In this paper, we proposed an approach integrating visualization and interactive techniques to facilitate multi-scale and cross-scale analysis in networks over affiliation hierarchy. The validity of this approach has been tested in exploring the co-authorship network of authors in Journal of the Association for Information Science and Technology (JASIST) restricting to the field ‘information retrieval’. The hidden network patterns across hierarchical attributes were revealed and discussed. This study provides a new perspective for visual analysis of network with hierarchical attributes.

Contact and Legal Notice · Contact Address:
Conference: iConference 2017
Conference Software - ConfTool Pro 2.6.115+TC
© 2001 - 2018 by H. Weinreich, Hamburg, Germany