Skip to main content

Über dieses Buch

This book constitutes the refereed proceedings of the 6th International Conference on Multidisciplinary Social Networks Research, MISNC 2019, held in Wenzhou, China, in August 2019.
The 15 full papers presented were carefully reviewed and selected from 37 submissions. The papers deal with the following topics: social network, social network analysis, data engineering, data mining, user behavior.



Supervised Deep Learning for Hierarchical Image Data Retrieval

The techniques of feature extraction and representation on image data have been significantly progressed in recent years due to the development of deep learning. With a large number of representative image features being extracted from ImageNet by convolutional neural networks, many object recognizing applications were successfully accomplished effectively. In this paper, two supervised image retrieval models for retrieving images with similar hierarchical concept are investigated and compared. First, image features are extracted by pre-trained VGG convolutional networks. Then, the supervised retrieval models are learned from a set of images with hierarchical concept labels. The experimental results show that the hash-based model generally is superior to classifier-based model both in F1 measure and MAP no matter what in coarse level or fine level of concept hierarchy.
Been-Chian Chien, Yueh-Chia Hsu, Ya-Yu Huang

Why Do People Back Crowdfunding Projects?

This study employs social cognitive theory as a theoretical foundation to empirically explore the influential antecedents of backing intention on crowdfunding platforms. We collected 221 valid samples from crowdfunding fan pages on Facebook in Taiwan, and the data were analyzed using the partial least squares (PLS) method. The results indicate that individual factors (i.e., backing self-efficacy, rewards, and empathy) and environmental factor (i.e., website service quality) all have a positive effect on behavior (i.e., backing intention). We conclude with a summary of theoretical implications and practical implications, along with an opportunity for IS researchers and practitioners to extend our knowledge of compelling future research possibilities.
Ying-Feng Kuo, Cathy S. Lin, Chung-Hsien Wu, Tsung-Hsun Tsai

DHUM: A Dynamic Human Urban Mobility Model for Smart Public Transport

Smart transportation helps people in the smart city to reach their destinations securely on time with the minimum waiting time. In general, the public transport system is scheduled based on different time of the day which not depends on other parameters like traffic, dynamic way of an understanding number of traveling passengers, etc. Due to this, some of the routes are heavily crowded and some are not. These can result in wastage of resources and not bring any smartness to public transport and it also causes discomfort to the citizen’s. In addition, this may lead to higher waiting times and force people to use their personal vehicles, ultimately, it increases the pollution level in cities. In this work, we propose a new method to schedule public transport by predicting the traveling passengers in each location. For this, we explore a way to understand the mobility patterns of citizen’s. DHUM, a dynamic human urban mobility model will be proposed for effective urban planning and transportation by exploring the use of mobile crowdsourcing data and heat maps in addition to their mobility pattern. This can help authorities to schedule public transport in real-time intelligently so that the citizen’s hassles may be governed to make them reach their destination on time. With the use of Dakar city data, we have rationalized that our proposed method can result in a decrease in pollution and subsequently it helps to increase the people’s comfort in future smart cities.
M. Saravanan, Perepu Satheesh Kumar

Effects of Consumption Value on Online Repurchase Intention: Mediation Effect of Green Information Visibility

The purpose of the paper is to propose and examine empirically a research model that describes the online repurchase intention. The proposed model is based on the concepts of expectation-confirmation theory and consumption value theory that includes internal, external, and functional values as the determinants of online consumer’s buying satisfaction towards online repurchasing intention. The green information visibility as a mediator to influence the effect of consumption values on consumer purchasing satisfaction is also examined. Based on the analysis of 308 valid samples, the main research findings are as follows. (1) Among those three values, internal and functional value highly influence their satisfaction. (2) Functional value is the only variable that is affected by the mediator while internal and external are not affected. (3) For sub-factors, quality and usability strongly support consumer’s satisfaction while price and attitude partially support their satisfaction. (4) Attitude, price and quality value has mediating effect except the others. Discussion and implications will also be delineated.
Jia-Hong Lim, Jung-Chen Chen, Chien Hsing Wu

An Effective BI-encoded Schema for Mention Extraction

We present a neural-encoded mention-hypergraph (named as NEMH in this paper) model for mention-extraction and classification in this paper. Through extraction of textual mention entities, a model is proposed that applies a hypergraph-encoding schema to neural networks. Comparing the results of the proposed model with the previous approaches, the proposed model can thus identify unlimited-length nested mention entities, which is a major milestone in the field. Several experiments are conducted on many datasets used in the baseline approaches, and the obtained results indicated that the designed model has high effectiveness compared to the existing models.
Jerry Chun-Wei Lin, Jimmy Ming-Tai Wu, Yinan Shao, Matin Pirouz, Binbin Zhang

A Latent Variable CRF Model for Labeling Prediction

A latent variable conditional random fields (CRF) model is proposed to improve sequence labeling, which utilizes the BIO encoding schema as latent variable to capture the latent structure of hidden variables and observation data. The proposed model automatically selects the best encoding schema for each given input sequence. Through experimentation, it is demonstrated that the proposed model unveils the latent variable while performing robustly on sequence-labeling prediction tasks.
Jerry Chun-Wei Lin, Jimmy Ming-Tai Wu, Yinan Shao, Matin Pirouz, Binbin Zhang

Exploring the Influencing Factors of Live-Streaming Viewers’ Participation Intention from the Perspective of Source Credibility Model and Cognitive Load - An Example of Mobile Device Users

In this study, we divide various behaviors of live-streaming viewers into non-creating, contributing and money-support three levels to exploring the influencing factors of live-streaming viewers’ participation intention. Streamers being the source of information in the virtual live-streaming room, their traits or the messages they transmitted deeply affect the viewers’ psychological state and behaviors that the viewers may act on the live-streaming platform. In addition, since the interactive mode of live-streaming program is that the streamer needs to face a large number of viewers in the same time, which means there is almost no direct interaction between the streamer and the viewers. The viewers only mostly imagine the interactive relationship between the two parties. In summary, we investigate the interaction of source credibility model, para-social interaction, cognitive load and the consequent outcome of participation intention for live-streaming platforms.
Shu-Chen Yang, Tzu-Ting Feng, Yu-Hui Wang

Post-purchase Dissonance of Mobile Games Consumer

Mobile game is the largest gaming platform nowadays. Most of mobile games adopt Freemium model selling virtual product in games as their revenue model. While research shows Freemium mobile games have an average life span of only ninety days, many players left the games soon after entering the game. One of factors that players left games is post-purchase dissonance. Post-purchase dissonance refers to a state that consumers feel regretful, frustrated or think they have made a wrong decision after the purchase. Prior research found that post-purchase dissonance will negatively influence satisfaction or lead to spread negative word-of-mouth.
In this study, we use expectation confirmation theory to examine which purchase motivation would cause greater dissonance on mobile game consumer. 18 in-game purchase motivations come from prior research and in-depth discussions with industry experts. The questionnaire is used to investigate on mobile games consumer in Taiwan.
Two-step cluster analysis was conducted and 18 in-game purchase motivations were divided into four cluster according to the level of post-purchase dissonance: Social & functional motivation, Affective value motivation, Impulsive buying motivation, and Speculative motivation. It is found that Speculative motivation and Impulsive buying motivation would cause greater post-purchase dissonance. Result of the research can help us understanding the antecedents of post-purchase dissonance on mobile games consumer and serve as a reference for game developers to design game mechanism.
Shu-Chen Yang, Rui-Min Chang, Chia-Jung Hsu

Japanese University Students’ Acceptance of Cross-border Electronic Commerce

With increasing smartphone penetration, online purchasing has become a common and representative purchasing channel. The development of the Internet, enabling online transactions, has led to the emergence of cross-border electronic commerce. Despite the fact that online shopping sites are developing very quickly, the factors that influence consumers’ purchase intention (PI) on cross-border online shopping sites are not clear. Furthermore, prior studies have not identified the perceived differences between domestic and cross-border online shopping sites.
This study examines young people’s acceptance of both domestic and cross-border online shopping sites, and identifies the relationships among the variables involved in online shopping sites. This study indicates that the PI with online shopping sites is directly influenced by perceived usefulness (PU), trust (TR), and credibility of personal information protection (PIP) on online shopping sites. In addition, this study reveals the perceived differences between domestic and cross-border online shopping sites. In particular, credibility of PIP and TR of cross-border online shopping sites strongly influences PI, and PU weakly influences PI. On the other hand, PU, PIP, and TR of domestic online shopping sites have a similar influence on PI.
Takashi Okamoto, Jiro Yatsuhashi, Naoki Mizutani

Visualization of Health Data

As data becomes more accessible, visualization methods are needed to help make sense of the information. Analyzing and visualizing data helps the public to better recognize the patterns and connections between different datasets. By using visual elements such as graphs, charts, and maps, it is easier to see and understand the trends and outliers in data. This project aims to study the correlation between environmental factors and public health. Large sets of data pertaining to the environment and health were gathered from open data sources. The tool used to analyze and visualize the collected data is Tableau, which is a software program that is used to transform data into dashboards and visuals such as treemaps, histograms, or area charts. For this project, the data will be displayed through charts and interactive maps that will be created through this software.
Veronica Castro Alvarez, Ching-yu Huang

Augmented Reality as a Reinforcement to Facilitate ESP Learning for Nursing Students

There is a great deal of research focusing on the benefit of teaching English for Specific Purpose (ESP) and yet there are relatively few studies suggesting ways how language teachers can help learners cope with learning by implementing Augmented Reality (AR). Therefore, the purpose of this study was to identify whether the integration of AR helped increasing EFL learners’ learning Effectiveness and motivation. The research took place in the nursing department at a private university in the northern part of Taiwan. 48 junior nursing students participated in the study. Mixed method was applied in the study. The questionnaire of AR assists learning L2 effectiveness and satisfaction was implemented at the first section of the data collection. In addition, the use of semi-structure interview and field note techniques were applied to be the main method of gathering qualitative research data. The data was obtained through adapted and modified open-ended questions. Participants were asked to reflect on how AR operated in the ESP context and how AR created the model of interactive learning environments after each AR learning practice. The results indicated that implementing AR was a vital strategy for producing independent thinkers and learners. They flourish under ESP-driven and creative approach; nursing students will be able to develop their own inquiries.
YingLing Chen

Improving Supply Chain Resilience with a Hybrid System Architecture

The purpose of this research is to propose a new hybrid system architecture, which is a solution for reducing and even avoiding the complexity and vulnerability of the whole supply chain. In this paper, we propose a conceptional model which introduces some features of decentralized system architecture based on blockchain technologies, such as new consensus mechanism and smart contract into a typical supply chain system so that defects of traditional centralized system will be solved and supply chain resilience can be improved. Furthermore, some fatal flaws of current blockchain system can be shored up in an IoT infrastructure environment which is consolidating supply chain resilience. Last but not lease, we identify some unsettled issues and drawbacks derived from the hybrid system architecture and indicate the solutions regarding these challenges.
Yu Cui, Hiroki Idota, Masaharu Ota

An Interactive, Location-Aware Taiwanese Social Network for Both Everyday Use and Disaster Management

The concept of leveraging social network crowdsourcing to help governments respond to and manage disaster events has become popular in recent years. However, existing methods in Taiwan have been challenged by the following difficulties. (1) Information crowdsourced from existing social networks has typically been limited by privacy settings (which can prevent data collection) and formatting issues (which can affect data processing). (2) Existing social networks lack location-aware functions, which makes it difficult to do the spatial analyses of disaster events. (3) Platforms which have been created to collect disaster information do not typically attract enough users to facilitate effective crowdsourcing of information following an disaster. To address these limitations, this study proposes a novel, interactive, and location-aware Taiwanese social network. Unlike social networks which are currently popular, our proposed network focuses on geographic space, and users can only view posts which were generated near their current location or close to where they live. Under normal, non-disaster circumstances, users can employ our proposed social network to view issues in their surrounding environment. In the event of an disaster, users can employ our social network to provide disaster management agencies with comprehensive information about the disaster. In addition, disaster management agencies can use the proposed social network to interact with users who are located near the disaster. We anticipate that our proposed social network will better facilitate information crowdsourcing than convention social networks in Taiwan.
Tzu-Ping Lai, Shun Yao, Wing-Lun Siu, Yu-Che Cheng, Heng-Yi Su, Yi-Chung Chen

A Fast Approach of Graph Embedding Using Broad Learning System

In this paper, traditional DeepWalk method and broad learning system (BLS) are used to classify network nodes in graph embedding, and results of classification are compared. When categorizing, DeepWalk adopts one vs rest (OvR) logistic regression method, and BLS is applied after the production of vector representations. In order to obviously compare results of the two classification methods, Support Vector Machine (SVM) and Convolutional Neural Network (CNN) are employed to carry out the experiment on multi-label classification of BlogCatalog. The experimental result shows that F1 score of BLS is obviously higher than DeepWalk and other methods, and training time of BLS is much less than other methods. These performances make our method suitable to graph embedding.
Long Jiang, Yi Zuo, Tieshan Li, C. L. Philip Chen

A Study on Pricing and Diffusion of New Services – an Experimental Research on Electronic Books Market

We treat distribution matters of Japanese electronic book service in this article. In this context the relationship between distribution factors and models are so important. In this paper we compare this issue with electronic money cards. Then review our research analysis from the view point of this prospect.
Masashi Ueda


Weitere Informationen

Premium Partner