SPAE: Lifelong disk failure prediction via end-to-end GAN-based anomaly detection with ensemble update
作者:
Liu, Yu;Guan, Yunchuan;Jiang, Tianming;Zhou, Ke;Wang, Hua;...
期刊:
Future Generation Computer Systems ,2023年148:460-471 ISSN:0167-739X
通讯作者:
Jiang, TM
作者机构:
[Liu, Yu; Zhou, Ke; Hu, Guangxing; Guan, Yunchuan; Wang, Hua; Jiang, Tianming] Huazhong Univ Sci & Technol, Wuhan Natl Lab Optoelect, Wuhan 430074, Peoples R China.;[Jiang, Tianming] Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.;[Liu, Yu] Huazhong Univ Sci & Technol, Sch Comp Sci & Technol, Wuhan 430074, Peoples R China.;[Cheng, Zhuo; Zhang, Ji; Fang, Wei] HUAWEI Technol, Innovat Ctr Data Storage, Chengdu 611730, Peoples R China.;[Huang, Ping] Temple Univ, Philadelphia, PA USA.
通讯机构:
[Jiang, TM ] ;Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.
关键词:
Adversarial training;Anomaly detection;Data reliability;Disk failure;Ensemble update;SMART
摘要:
Disk failure prediction aims to predict upcoming disk failures in advance for high data reliability. There are numerous supervised machine learning methods that are successful in predicting disk failure using SMART properties as input. However, these approaches heavily rely on a substantial number of annotated failed disks, resulting in degraded prediction performance caused by scarce failed disks at the beginning, also known as the cold start problem. Inspired by the success achieved in Generative Adversarial Network (GAN) based anomaly detection, this paper translates disk failure prediction into an anomaly detection problem. Specifically, we developed a Semi-supervised method for lifelong disk failure Prediction via Adversarial training and Ensemble update, called SPAE. The advantage of SPAE over existing supervised approaches is that SPAE can train the prediction model using only healthy disks, avoiding the cold start problem. Furthermore, SPAE can be updated using ensemble learning on emerging failed disks to resist the model aging problem. Compared to state-of-the-art methods using supervised machine learning on real-world datasets, SPAE predicts disk failures with higher accuracy for the full lifetime of models, i.e., both the startup period and the long-term usage. © 2023 Elsevier B.V.
语种:
英文
展开
Robots’ picking efficiency and pickers’ energy expenditure: the item storage assignment policy in robotic mobile fulfillment system
作者:
Zhang, Jun;Zhang, Ning;Tian, Lingkun;Zhou, Zijuan;Wang, Ping
期刊:
Computers & Industrial Engineering ,2023年176:108918 ISSN:0360-8352
通讯作者:
Zhang, Jun(zhangj@ccnu.edu.cn)
作者机构:
[Tian, Lingkun; Zhou, Zijuan; Wang, Ping; Zhang, Jun; Zhang, Ning] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Wang, Ping; Zhang, Jun] Cent China Normal Univ, Ecommerce Res Ctr Hubei Prov, Wuhan, Peoples R China.
通讯机构:
[Zhang, J.] S;School of Information Management, China
关键词:
Bi-objective optimization problem;Energy expenditure;Human-robot coordinated;Multi-objective evolutionary algorithms;Robotic mobile fulfilment system;Storage assignment problem
摘要:
The item storage assignment problem (ISAP) in a robotic mobile fulfillment system (RMFS) is addressed in the paper. Recently, most ISAP studies have concentrated on improving the robots’ picking efficiency while ignoring the fact that RMFS is a human-robot coordinated system. In ISAP, we also need to take human factors into account. This research investigates ISAP by considering both the robots’ picking efficiency and the pickers’ energy expenditure. The bi-objective mixed-integer programming model is developed to maximize the items’ similarities in the pods and minimize the pickers’ energy expenditure. The improved knee point-driven evolution algorithm (IKnEA) is proposed to solve the model. Experimental results demonstrate that IKnEA has superior distribution and convergence speed than NSGA-II and KnEA. IKnEA's hypervolume (HV) indicator operates better than NSGA-II and KnEA in various warehouse scales. More layers of pods contribute to less travel distance for robots but more energy expenditure for pickers. Five Pareto solutions are presented to assist the decision-makers in selecting the optimal ISAP solutions given various preferences. © 2022 Elsevier Ltd
语种:
英文
展开
Exploring the influencing factors of urban residential electricity consumption in China
作者:
Hao, Peng;Yin, Shimin;Wang, Dan;Wang, Jiancheng
期刊:
Energy for Sustainable Development ,2023年72:278-289 ISSN:0973-0826
通讯作者:
Wang, Jiancheng(wangjc1987@foxmail.com)
作者机构:
[Yin, Shimin; Hao, Peng] Anhui Univ, Sch Business, Hefei 230039, Peoples R China.;[Wang, Dan] Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.;[Wang, Jiancheng] Wuhan Univ Technol, Sch Econ, Wuhan 430070, Peoples R China.
通讯机构:
[Jiancheng Wang] S;School of Economics, Wuhan University of Technology, Wuhan 430070, China
关键词:
Electricity consumption;Urban residents;GDIM;China
摘要:
China has witnessed 556 million people migrating to urban areas over the past three decades with the process of urbanization, which has brought enormous pressure on electricity consumption and urban sustainable development. This paper establishes the generalized Divisia Index Decomposition Method (GDIM) model, and aims to reveal the electricity consumption mechanism of urban residents in China between 2000 and 2018. The results show that residential income, per capita electricity consumption and residential appliances are the main driving factors for the growth of urban residential electricity consumption, while the energy-saving willingness and expenditure structure have the inhibiting effect. With the expansion of urban residential income, the increase in income and air-conditioning numbers will continue to promote the growth of urban residential electricity consumption. Then, this paper puts forward relevant suggestions to alleviate the pressure of electricity demand and sustainable development from urban residents in China. © 2023 International Energy Initiative
语种:
英文
展开
Cross-modal dynamic sentiment annotation for speech sentiment analysis
作者:
Chen, Jincai;Sun, Chao;Zhang, Sheng;Zeng, Jiangfeng
期刊:
Computers & Electrical Engineering ,2023年106:108598 ISSN:0045-7906
通讯作者:
Zeng, Jiangfeng(jfzeng@ccnu.edu.cn)
作者机构:
[Zhang, Sheng; Chen, Jincai; Sun, Chao] Huazhong Univ Sci & Technol, Wuhan Natl Lab Optoelect, Wuhan 430074, Peoples R China.;[Zeng, Jiangfeng] Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.
通讯机构:
[Jiangfeng Zeng] S;School of Information Management, Central China Normal University, Wuhan 430079, China
关键词:
Cross-modal annotation;Multi-modal video;Sentiment profiles;Speech sentiment analysis
摘要:
Traditionally, one single hard label determines the sentiment label of an entire utterance for speech sentiment analysis. It obviously ignores the inherent dynamic and ambiguity of speech sentiments. Moreover, there are few segment-level ground truth labels in the most existing sentiment corpora, due to the label ambiguity and annotation cost. In this work, to capture segment-level sentiment fluctuations across one utterance, we propose sentiment profiles (SPs) to express segment-level soft labels. Meanwhile, we introduce massive multi-modal wild video data to solve the data shortage problem, and facial expression knowledge is used to guide audio segments generate soft labels through the Cross-modal Sentiment Annotation Module. Then, we design a Speech Encoder Module to encode audio segments into SPs. We further exploit the sentiment profile purifier (SPP) to iteratively improve the accuracy of SPs. Numerous experiments show that our model achieves state-of-the-art performance on CH-SIMS and IEMOCAP datasets with unlabeled data respectively. © 2023 Elsevier Ltd
语种:
英文
展开
Green Supplier Selection Mechanism Based on Information Environment of Z-Numbers
作者:
Rao, Congjun;Gao, Mingyun;Goh, Mark;Xiao, Xinping
期刊:
Cognitive Computation ,2023年15(2):520-533 ISSN:1866-9956
通讯作者:
Gao, Mingyun(gaomyhnu@163.com)
作者机构:
[Rao, Congjun; Xiao, Xinping] Wuhan Univ Technol, Sch Sci, Wuhan 430070, Peoples R China.;[Gao, Mingyun] Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.;[Goh, Mark] Natl Univ Singapore, NUS Business Sch, Singapore 119623, Singapore.;[Goh, Mark] Natl Univ Singapore, Logist Inst Asia Pacific, Singapore 119623, Singapore.
通讯机构:
[Mingyun Gao] S;School of Information Management, Central China Normal University, Wuhan, People’s Republic of China
关键词:
Multi-attribute decision making;Green supplier selection;Z-numbers;Possibility degree
摘要:
In the green procurement management of firms, by rigorously evaluating and selecting suppliers, firms can effectively reduce the operation risk and improve the sustainable value-added capacity of supply chain. This paper designs a novel green supplier selection mechanism based on an uncertain information environment. First, a new green supplier evaluation index system is designed by considering the economic, environmental, and social drivers from the perspective of sustainable development. Then, the Z-numbers are applied to describe the fuzziness degree of the decision information and the reliability degree of the fuzzy attribute values. We re-define the possibility degree of trapezoidal fuzzy numbers, and on that basis, we define the possibility degree of Z-numbers. A ranking method based on the possibility degree of Z-numbers is applied to select the green suppliers. Finally, the implementation, applicability, and feasibility of the proposed mechanism are highlighted by providing a decision-making example of green supply selection together with the comparison analysis with the existing methods. From the comparison analysis and discussion, the information processing method in our proposed mechanism can effectively avoid the information loss caused by the direct aggregation of fuzzy information, which shows that the proposed mechanism is more feasible and effective than other congeneric methods. The results suggest that the proposed mechanism of green supplier selection can handle multi-attribute decision-making problems in an uncertain cognitive information environment with Z-numbers, consistent with human cognition. © 2022, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.
语种:
英文
展开
Considering pickers’ learning effects in selecting between batch picking and batch-synchronized zone picking for online-to-offline groceries
作者:
Zhang, Jun;Zhang, Xueyan;Zhang, Ning
期刊:
Applied Mathematical Modelling ,2023年113:358-375 ISSN:0307-904X
通讯作者:
Zhang, Jun(zhangj@ccnu.edu.cn)
作者机构:
[Zhang, Jun; Zhang, Xueyan; Zhang, Ning] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.
通讯机构:
[Jun Zhang] S;School of Information Management, Central China Normal University, Wuhan, China
关键词:
Batch picking;Batch-synchronized zone picking;Learning effects;Online-to-offline grocery
摘要:
The performance of two common picking strategies, batch picking (BP) and batch-synchronized zone picking (BSZP), are compared by considering pickers’ learning effects in online-to-offline (O2O) groceries. As we know from previous research, one major advantage of BSZP is that pickers can be more familiar with the item locations because each picker works in a relatively smaller picking area. However, few papers have formulated a quantitative method to verify this advantage. To address this, we use pickers’ learning effects to quantify their familiarity with the item locations, and then compare the picking efficiency in BP and BSZP with learning effects. The optimization models of BP and BSZP are formulated to minimize the total service time, and heuristic methods are proposed to solve these models. The case study is implemented under three order types and three storage assignment policies, i.e. Random, ABC-1, and ABC-2. The results of the study indicate that with learning effects, although BSZP needs a longer travel time than BP, it requires a shorter search time. Which strategy, BP or BSZP, is more efficient mainly depends on the learning rates, storage assignment policies, and pickers’ working days. When the storage assignments are Random and ABC-1, a smaller learning rate can lead BSZP to perform better than BP. When the storage assignment is ABC-2, BSZP performs better than BP. When pickers’ working days are long enough, the learning effect is weakened, and BP outperforms. Awareness of the learning effects is required to choose a better picking strategy, evaluate more accurate picking efficiency and hire an appropriate number of pickers. © 2022
语种:
英文
展开
Multi-variety and small-batch production quality forecasting by novel data-driven grey Weibull model
作者:
Xiao, Qinzi;Gao, Mingyun;Chen, Lin;Goh, Mark
期刊:
Engineering Applications of Artificial Intelligence ,2023年125:106725 ISSN:0952-1976
通讯作者:
Gao, MY
作者机构:
[Xiao, Qinzi; Chen, Lin] Wuhan Inst Technol, Sch Management, Wuhan 430205, Peoples R China.;[Gao, Mingyun] Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.;[Goh, Mark] Natl Univ Singapore, NUS Business Sch, Singapore 117592, Singapore.;[Goh, Mark] Natl Univ Singapore, Logist Inst Asia Pacific, Singapore 117592, Singapore.
通讯机构:
[Gao, MY ] C;Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.
关键词:
Data-driven;Grey system;Grey Weibull model;Multi-variety and small-batch production;Product quality prediction
摘要:
With the coming of intelligent manufacturing, multi-variety and small-batch production mode has gradually become popular. Aiming at the characteristics of high dimensional information and limited samples in this mode, a novel data-driven grey Weibull model is established for product quality prediction. Firstly, a high-dimensional information from the production process is integrated as the process variation using a data-driven method. Then, the quality prediction function is deducted by considering that process variation follows Weibull distribution-based mechanism, the Hausdorff difference scheme is adopted to weaken the error from the difference to the differential, heuristic algorithm is selected to optimize the distribution parameters of the model. Finally, an experimental analysis is designed using the dataset from some personalized customization manufacturer in China. Results show that the proposed model is not only superior to the other eight models in terms of stability and prediction accuracy, but also boasts the features of amalgamation of data-driven and mechanism-driven methods, which can simultaneously process high-dimensional information and limited samples in multi-variety and small-batch production system. © 2023 Elsevier Ltd
语种:
英文
展开
Monopoly or competition: strategic analysis of a retailing technology service provision
作者:
Hu, Fengying;Zhou, Zhenglong*
期刊:
Electronic Commerce Research ,2022年22(4):1651-1689 ISSN:1389-5753
通讯作者:
Zhou, Zhenglong
作者机构:
[Hu, Fengying] Hubei Univ, Business Sch, Wuhan 430062, Peoples R China.;[Zhou, Zhenglong] Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.
通讯机构:
[Zhou, Zhenglong] C;Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.
关键词:
Retailing technology service;Strategic choice;Retail competition;Power structure
摘要:
To improve the consumer shopping experience and to remain competitive, an increasing number of retailers continue to increase their investment in technology services and actively seek help from technology service providers to transform and upgrade their retail business. Technology service providers deliver new technological solutions for retailers by providing professional retailing technology services (e.g., building omnichannel retailing modes). Considering this context, we seek to understand whether retailers expect to use technology services and service providers’ expectations of providing retailing services. In this paper, we build a two-tier supply chain with one upstream technology service provider and one downstream retailer (or two competitive retailers) to analyse the strategic choice of the retailer(s) and the service provider. We find that the retailer(s) can purchase a professional retailing technology from the technology service provider and that the service provider is willing to provide technology to two competing retailers, which can increase market demand and yield higher profits. Furthermore, we findthe impact of the technological contribution level, retail competition intensity, the power structure, and the technology marketing effect on decision outcomes issignificant. Therefore, retailers should consider the influence of these factors when service providers deliver a retailing technology service. © 2020, Springer Science+Business Media, LLC, part of Springer Nature.
语种:
英文
展开
Disclosure decisions and the moderating effects of privacy feedback and choice
作者:
Liu, Bailing;Miltgen, Caroline Lancelot;Xia, Huimin
期刊:
Decision Support Systems ,2022年155:113717 ISSN:0167-9236
通讯作者:
Miltgen, Caroline Lancelot(clancelot@audencia.com)
作者机构:
[Liu, Bailing] Shanghai Int Studies Univ, Sch Business & Management, Lab Appl Brain & Cognit Sci, Shanghai, Peoples R China.;[Liu, Bailing; Xia, Huimin] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Miltgen, Caroline Lancelot] Audencia Business Sch, 8 Route Joneliere, F-44312 Nantes, France.
通讯机构:
[Miltgen, C.L.] A;Audencia Business School, 8 Route de la Jonelière, Nantes, France
关键词:
Disclosure;Justice theory;Perceived privacy;Privacy feedback;Privacy protection
摘要:
To address data concealment challenges, this study considers how perceived privacy and justice might drive users' information disclosure intentions. Two contextual factors (privacy feedback and choice) also might function as moderators in the proposed relationships. The model was tested using a controlled lab experiment featuring a new app dedicated to m-commerce where both privacy feedback and choice were manipulated. The results affirm a critical role of perceived justice in determining perceived privacy, which shapes disclosure intentions. Providing privacy feedback also enhances the positive effect of perceived justice on perceived privacy and the positive effect of trust propensity on disclosure intention, and alleviates the negative effect of privacy experience on perceived privacy. In addition, providing choices can help users make more conscious privacy decisions. Accordingly, this study contributes to extant literature by extending a single-context theory view and exploring moderating effects of both privacy feedback and choice on information disclosure decision processes. This research also establishes the real effect of feedback by differentiating people who choose (or not) to receive privacy feedback. In addition, this study is applying justice theory to design privacy feedback, through a theoretical link between feedback functions and three justice dimensions (i.e., procedural, informational, and interactional). Finally, we replicate previous findings from past privacy literature using a new setting and a new measure for privacy, further establishing confidence in those relationships. © 2022 Elsevier B.V.
语种:
英文
展开
Mapping the evolution of e-commerce research through co-word analysis: 2001–2020
作者:
Bai, Yang;Li, Hongxiu
期刊:
Electronic Commerce Research and Applications ,2022年55:101190 ISSN:1567-4223
通讯作者:
Li, Hongxiu(Hongxiu.li@tuni.fi)
作者机构:
[Bai, Yang] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Li, Hongxiu] Tampere Univ, Fac Management & Business, Dept Informat & Knowledge Management, Tampere, Finland.
通讯机构:
[Li, H.] D;Department of Information and Knowledge Management, Finland
关键词:
Co-word analysis;E-commerce, research trends;Research evolution;Research theme
摘要:
This current study examines how research themes and trends have developed in the e-commerce field, aiming to provide a comprehensive understanding of e-commerce research based on prior literature. To answer this question, the current study applied a bibliometric method, word co-occurrence (co-word) analysis, to identify core research themes and trends in the e-commerce field, based on 17,416 author-provided keywords collected from 3,284 academic articles published between 2001 and 2020 in seven major journals in e-commerce field. Specifically, the data were analyzed based on two periods: 2001–2010 and 2011–2020. We found that the research themes in the e-commerce field have evolved alongside technology's development and diffusion. Some research themes, such as recommended system and eGovernment, have persisted as important research themes and studied continuously over the 20 years studied. Some research themes studied in 2001–2010, such as B2C and XML, disappeared in 2011–2020, whereas some new research topics, such as cloud computing, mobile technologies, artificial intelligence, and machine learning, emerged in recent years. This study provides a good understanding of e-commerce research's evolution, as well as useful suggestions to direct future research by scholars in the e-commerce field. © 2022 The Author(s)
语种:
英文
展开
Model-free finite-horizon optimal tracking control of discrete-time linear systems
作者:
Wang, Wei;Xie, Xiangpeng;Feng, Changyang
期刊:
Applied Mathematics and Computation ,2022年433:127400 ISSN:0096-3003
通讯作者:
Feng, Changyang(cyfeng@ccnu.edu.cn)
作者机构:
[Wang, Wei] Zhongnan Univ Econ & Law, Sch Informat & Safety Engn, Wuhan 430073, Peoples R China.;[Wang, Wei] Hubei Key Lab Adv Control & Intelligent Automation, Wuhan, Peoples R China.;[Wang, Wei] Minist Educ, Engn Res Ctr Intelligent Technol Geoexplorat, Wuhan, Peoples R China.;[Xie, Xiangpeng] Nanjing Univ Posts & Telecommun, Inst Adv Technol, 210 0 03, Nanjing, Peoples R China.;[Feng, Changyang] Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.
通讯机构:
[Changyang Feng] S;School of Information Management, Central China Normal University, Wuhan 430079, PR China
关键词:
Q-function;Finite-horizon;Linear quadratic tracking;Model-free
摘要:
Conventionally, the finite-horizon linear quadratic tracking (FHLQT) problem relies on solving the time-varying Riccati equations and the time-varying non-causal difference equations as the system dynamics is known. In this paper, with unknown system dynamics being considered, a Q-function-based model-free method is developed to solve the FHLQT problem. First, an augmented system consisting of the controlled system and the desired trajectory system is formulated, and the FHLQT problem transforms to the finite-horizon linear quadratic regulator (FHLQR) problem with the augmented system. Then, a time-varying Q-function which depends explicitly on the control input is defined. With the defined time-varying Q-function, a model-free finite-horizon control method is developed to approximate the solutions of the time-varying Riccati equations of the transformed FHLQR problem. At last, simulation studies are carried out to verify the validity of the developed method. © 2022 Elsevier Inc.
语种:
英文
展开
Users' mental models of cross-device search under controlled and autonomous motivations
作者:
Dong, Jing;Duan, Ruoyang;Liang, Shaobo
期刊:
Aslib Journal of Information Management ,2022年75(1):68-89 ISSN:2050-3806
通讯作者:
Dong, Jing(dongjing@ccnu.edu.cn)
作者机构:
[Dong, Jing; Duan, Ruoyang] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Liang, Shaobo] Wuhan Univ, Sch Informat Management, Wuhan, Peoples R China.
通讯机构:
[Dong, J.] S;School of Information Management, Central China Normal University, Wuhan, China
关键词:
Cross-device search;Mental models;Controlled motivation;Autonomous motivation;Self-determination theory;Grounded theory
摘要:
Purpose: Existing literature has elicited the external behaviors of cross-device search but not much is known about users' cognition of cross-device search. The study aims to explore how users perceive the cross-device search by combining with the mental models and how motivations affect the users' understanding of cross-device search. Design/methodology/approach: The research questions are addressed through semi-structured face-to-face interviews with 59 users. Prior to the interview, the user is asked to complete a cross-device search task designed with the simulation of controlled and autonomous motivations to gain a real experience. The concepts of mental models are coded according to the constructivist grounded theory method. Findings: The study finds the users' mental models of cross-device search consist of four dimensions: Element, Quality, Function and Issue. The effect of motivation on the mental models is tested as significant in terms of the Quality and Function aspects. The controlled motivation affects the user's perception of how the device switch influence the search and the autonomous motivation influences the user's opinion of search system functions. Originality/value: The contribution of this study is found to extend the existing knowledge of cross-device search and update the mental models of information search in the current multi-device environment. The findings inform the future study of cross-device search and practices of search system design. © 2022, Emerald Publishing Limited.
语种:
英文
展开
Cross-modal fusion for multi-label image classification with attention mechanism
作者:
Wang, Yangtao;Xie, Yanzhao;Zeng, Jiangfeng;Wang, Hanpin;Fan, Lisheng;...
期刊:
Computers & Electrical Engineering ,2022年101:108002 ISSN:0045-7906
通讯作者:
Zeng, Jiangfeng(jfzeng@mail.ccnu.edu.cn)
作者机构:
[Wang, Hanpin; Wang, Yangtao; Fan, Lisheng] Guangzhou Univ, Sch Comp Sci & Cyber Engn, Guangzhou, Peoples R China.;[Xie, Yanzhao] Huazhong Univ Sci & Technol, Wuhan Natl Lab Optoelect, Wuhan, Peoples R China.;[Zeng, Jiangfeng] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Song, Yufan] Nanjing Univ Posts & Telecommun, Nanjing, Peoples R China.
通讯机构:
[Jiangfeng Zeng] S;School of Information Management, Central China Normal University, China
关键词:
Attention mechanism;Cross-modal fusion;Graph convolution network;Multi-label image classification
摘要:
For multi-label image classification, existing studies either utilize a poor multi-step training workflow to explore the (local) relationships between the image target regions and their corresponding labels with attention mechanism or model the (global) label dependencies via graph convolution network (GCN) but fail to efficiently fuse these image features and label word vectors. To address these problems, we develop Cross-modal Fusion for Multi-label Image Classification with attention mechanism (termed as CFMIC), which combines attention mechanism and GCN to capture the local and global label dependencies simultaneously in an end-to-end manner. CFMIC mainly contains three key modules: (1) a feature extraction module with attention mechanism which helps generate the accurate feature of each input image by focusing on the relationships between image labels and image target regions, (2) a label co-occurrence embedding learning module with GCN which utilizes GCN to learn the relationships between different objects to generate the label co-occurrence embeddings and (3) a cross-modal fusion module with Multi-modal Factorized Bilinear pooling (termed as MFB) which efficiently fuses the above image features and label co-occurrence embeddings. Extensive experiments on MS-COCO and VOC2007 verify CFMIC greatly promotes the convergence efficiency and produces better classification results than the state-of-the-art approaches. © 2022 Elsevier Ltd
语种:
英文
展开
Hierarchical quantization for billion-scale similarity retrieval on GPUs
作者:
Chen, Wei;Ma, Xiao;Zeng, Jiangfeng;Duan, Yaoqing;Zhong, Grace
期刊:
Computers & Electrical Engineering ,2021年90:107002- ISSN:0045-7906
通讯作者:
Zeng, Jiangfeng(jfzeng@mail.ccnu.edu.cn)
作者机构:
[Chen, Wei] Hubei Normal Univ, Coll Comp & Informat Engn, Huangshi, Hubei, Peoples R China.;[Ma, Xiao] Zhongnan Univ Econ & Law, Sch Informat & Safety Engn, Wuhan, Peoples R China.;[Duan, Yaoqing; Zeng, Jiangfeng] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Zhong, Grace] China Merchants Huihe Equ Fund Management Co Ltd, Shenzhen, Peoples R China.
通讯机构:
[Zeng, J.] S;School of Information Management, China
关键词:
Approximate nearest neighbor search;Billion-scale high-dimensional data;GPU;Quantization
摘要:
Existing billion-scale approximate nearest neighbor (ANN) search methods are usually based on GPU implementations. However, the search accuracy of these methods are limited by their indexing structures and encoding algorithms which are not adaptable to GPUs. We introduce an ANN search method called Vector and Product Quantization Tree (V-PQT) based on GPUs, based on a novel two-level tree indexing structure that increase indexing efficiency and a novel encoding algorithm that improved encoding quality of previous quantization methods by a large margin. Comparing to other state-of-the-art methods based on GPUs, V-PQT system can provide a significantly better recall with the same level of runtime. The experimental result shows that the proposed method has 27% and 20% improved to the conventional methods in the term of search accuracy on two public billion-scale datasets. © 2021 Elsevier Ltd
语种:
英文
展开
HMM-BiMM: Hidden Markov Model-based word segmentation via improved Bi-directional Maximal Matching algorithm
作者:
Yan, Xingyu;Xiong, Xiaofan;Cheng, Xiufeng;Huang, Yujing;Zhu, Haitao;...
期刊:
Computers & Electrical Engineering ,2021年94:107354- ISSN:0045-7906
通讯作者:
Hu, Fang(naomifang@mails.ccnu.edu.cn)
作者机构:
[Hu, Fang; Huang, Yujing; Yan, Xingyu; Xiong, Xiaofan; Zhu, Haitao] Hubei Univ Chinese Med, Coll Informat Engn, Wuhan 430065, Peoples R China.;[Cheng, Xiufeng] Cent China Normal Univ, Sch Informat Management, Wuhan 430079, Peoples R China.;[Hu, Fang] Univ West Florida, Dept Math & Stat, Pensacola, FL 32514 USA.
通讯机构:
[Hu, F.] C;College of Information Engineering, China
关键词:
Bidirectional Maximal Matching;Dictionary dynamic update;Hidden Markov model;Medical text segmentation
摘要:
Combining with the Hidden Markov Model and Bi-directional Maximal Matching algorithm, a new word segmentation algorithm, HMM-BiMM, is presented. In terms of the sub-dictionary matching, it can implement a fast word segmentation. After segmenting the text by the Bi-directional Maximal Matching (BiMM), the remaining text connected by the remaining single words will be segmented again by the strategy of the Hidden Markov Model (HMM). By the HMM, this algorithm can realize the dictionary dynamic update by the new segmentation words and improve the segmentation accuracy accordingly. Compared with five representative algorithms in the real-world clinical text (symptom), we show that the HMM-BiMM algorithm achieves the highest efficiency and accuracy for symptom text segmentation. In detail, this algorithm has around 3% in precision and 70% in running time improved to the BiMM. © 2021 Elsevier Ltd
语种:
英文
展开
How privacy concerns and cultural differences affect public opinion during the COVID-19 pandemic: a case study
作者:
Xing, Yunfei;Li, Yuhai;Wang, Feng-Kwei
期刊:
Aslib Journal of Information Management ,2021年73(4):517-542 ISSN:2050-3806
通讯作者:
Xing, Yunfei(xingyf@mail.ccnu.edu.cn)
作者机构:
[Xing, Yunfei; Li, Yuhai] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Wang, Feng-Kwei] Tunghai Univ, Coll Management, Execut Master Program Business Adm, Taichung, Taiwan.
通讯机构:
[Xing, Y.] S;School of Information Management, China
关键词:
Privacy concern;National culture;Public opinion;National policy;COVID-19;Twitter;Weibo
摘要:
Purpose: COVID-19, an infectious disease first identified in China, has resulted in an ongoing pandemic all over the world. Most of the countries have been experiencing a difficult period during the fighting of this pandemic. The purpose of this study is to explore the effect of privacy concerns and cultural differences on public opinion related to the pandemic. The authors conducted a comparative analysis of public opinion in the US and in China as a case study, in order to determine the results. Design/methodology/approach: National policies on important issues faced during the COVID-19 pandemic in the US and in China were examined through a comparative analysis. The authors used text clustering and visualization to mine public opinion on two popular social media platforms, Twitter and Weibo. From the perspectives of concern for privacy and of national culture, this study combines qualitative and quantitative analysis to discover the acceptance level of national policies by the public in the two countries. Findings: The anti-pandemic policies and measures of the US and China reflect the different characteristics of their respective political systems and national cultures. When considering the culture of the US, it is hard to establish and enforce a rigorous regulation on either mask wearing in public or home quarantine on the national level. The opinions of US people are diverse, regarding national COVID-19 policies, but they are rather unified on privacy issues. On the other hand, Chinese people show a high acceptance of national policies based on their mask-wearing customs and their culture of collectivism. Originality/value: Prior studies have paid insufficient attention to the ways in which user privacy and cultural difference affect public opinion on national policies between the US and China. This case study that compares public opinion on current and topical issues which are closely bound up with public life shows originality, as it innovatively provides a cross-cultural perspective on the research of public opinion dissemination during emergencies by considering the ongoing COVID-19 pandemic. © 2021, Emerald Publishing Limited.
语种:
英文
展开
Information needs and services for autism in China: is there any gap between them?
作者:
Lu, Quan;Song, Bixuan;Chen, Jing;Xie, Iris;Shen, Yutian
期刊:
Aslib Journal of Information Management ,2021年73(4):497-516 ISSN:2050-3806
通讯作者:
Shen, Yutian(yutian@whu.edu.cn)
作者机构:
[Shen, Yutian; Song, Bixuan; Lu, Quan] Wuhan Univ, Ctr Studies Informat Resources, Wuhan, Peoples R China.;[Lu, Quan] Wuhan Univ, Big Data Inst, Wuhan, Peoples R China.;[Chen, Jing] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Xie, Iris] Univ Wisconsin, Sch Informat Studies, Milwaukee, WI 53201 USA.
通讯机构:
[Shen, Y.] C;Center for Studies of Information Resources, China
关键词:
Autism;Information need;Information service;Information gap;China;Multi-source data
摘要:
Purpose: This study aims to explore the gap between information needs and services for autism in China. Design/methodology/approach: The gap is revealed by investigating the status quo of autism information needs and services. The authors extracted categories and subcategories of information needs through content analysis of academic documents and then supplemented the subcategories through text mining of an online forum. Meanwhile, categories and subcategories of information services were extracted through content analysis of autism websites. Finally, the authors matched the two to explore the gap and designed a quantitative index to measure it. Findings: A total of eight and ten categories of information needs and services are extracted, respectively. In total, six categories of information needs can be partly matched, but nearly half of the subcategories failed. Huge gaps in economic support, sociality and policy mechanisms categories are observed through the quantitative index and medium gaps in social resource services and employment categories while almost no gaps in psychological/emotional counseling, rehabilitation skills training and professional knowledge/information categories. Originality/value: This study takes a deep insight into the gap between autism information needs and services in China, providing evidences and suggestions for information providers to improve their services. Academic documents and online forum data are adopted to avoid the impact of stigmatization, which provides a multi-source data analysis approach for the information needs of special groups. © 2021, Emerald Publishing Limited.
语种:
英文
展开
Collaboration in demand-driven supply chain: Based on a perspective of governance and IT-business strategic alignment
作者:
Chi, Maomao* ;Huang, Rui;George, Joey F.
期刊:
International Journal of Information Management ,2020年52:102062- ISSN:0268-4012
通讯作者:
Chi, Maomao
作者机构:
[Chi, Maomao] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Huang, Rui] Univ Massachusetts Dartmouth, Charlton Coll Business, Dartmouth, MA USA.;[George, Joey F.] Iowa State Univ, Ivy Coll Business, Ames, IA USA.
通讯机构:
[Chi, Maomao] C;Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.
关键词:
Formal contract;E-business strategic alignment;Supply chain collaboration;Degree of competition
摘要:
In today's marketplace, firms participating in a supply chain are required to work collaboratively to quickly respond to market and consumer demands. To understand how firms may effectively achieve supply chain collaboration, this paper presents an empirical study using a moderated mediation model that explores the effects of formal contract, e-business strategic alignment (including intellectual and operational alignment) and the degree of competition on collaboration in demand-driven supply chain. Using data from a survey of IT and business executives in 209 firms, we discover that formal contract has a positive effect on supply chain collaboration through intellectual and operational alignment. In addition, the positive mediating effect of operational alignment is stronger under a higher degree of competition. Although there is no moderated mediation effect in the path of formal contract-intellectual alignment-supply chain collaboration, the degree of competition shows a negative moderated effect on the relationship between formal contract and intellectual alignment, and a positive moderated effect on the relationship between intellectual alignment and supply chain collaboration. Learnings from this study contribute to supply chain management literature, and provide a comprehensive view of the governance and strategic antecedents of supply chain collaboration under various environmental circumstances. © 2020 Elsevier Ltd
语种:
英文
展开
Understanding verified users' posting behavior from the perspective of human dynamics: a case study of Sina micro-blog
作者:
Yi, Ming;Lu, Yingying;Deng, Weihua* ;Kun, Lu;Zhang, Zhanhao
期刊:
Aslib Journal of Information Management ,2020年73(2):221-239 ISSN:2050-3806
通讯作者:
Deng, Weihua
作者机构:
[Yi, Ming; Zhang, Zhanhao; Lu, Yingying] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.;[Deng, Weihua] Huazhong Agr Univ, Coll Publ Adm, Wuhan, Peoples R China.;[Kun, Lu] Univ Oklahoma, Sch Lib & Informat Studies, Norman, OK 73019 USA.
通讯机构:
[Deng, Weihua] H;Huazhong Agr Univ, Coll Publ Adm, Wuhan, Peoples R China.
关键词:
Micro-blog;Posting behavior;Task-driven model;Verified users;Human dynamics;Social media
摘要:
Purpose: The purpose of this paper is to present a new human dynamics model to explain the process of verified users' (VUs) posting on Sina micro-blog. Design/methodology/approach: A common human dynamics research method with three steps is applied. Firstly, a large-scale behavioral dataset is collected involving 495 VUs and five topics on Sina micro-blog. Second, five important indicators that reveal the characteristics of posting behavior are analyzed. Then, a quantitative model is constructed to describe the process of posting behavior, and its validity is verified by simulations. Findings: Three important characteristics of VUs' micro-blog posting behavior are observed: fat-tailed distribution, fluctuation and periodicity. These characteristics do not fit the assumption of interest-driven models proposed by previous literature. An optimized task-driven model is introduced to describe this complex phenomenon mathematically. The model is verified on empirical data, confirming that task-driven models can be optimized to explore information behavior on social media. Originality/value: Being different from previous studies that mainly describe common users' posting behavior on social media by applying interest-driven models, this paper customizes an optimized task-driven model for VUs, who mainly treat social media as a platform for work and play a crucial role in information creation on social media. © 2020, Emerald Publishing Limited.
语种:
英文
展开
Understanding co-editing mechanism of wiki-based digital humanities projects
作者:
Deng, Weihua;Lv, Pei;Yi, Ming* ;Liu, Ming
期刊:
Aslib Journal of Information Management ,2020年72(2):199-218 ISSN:2050-3806
通讯作者:
Yi, Ming
作者机构:
[Deng, Weihua; Lv, Pei] Huazhong Agr Univ, Coll Publ Adm, Dept Informat Management, Wuhan, Peoples R China.;[Liu, Ming; Yi, Ming] Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.
通讯机构:
[Yi, Ming] C;Cent China Normal Univ, Sch Informat Management, Wuhan, Peoples R China.
关键词:
Crowdsourcing;Digital humanities;Consensus;Co-editing;Culture heritage;IBIS
摘要:
Purpose: The purpose of this paper is to reveal the co-editing mechanism aiming at content creation, and an entry of online encyclopedia is taken as a case, for the purpose of promoting and enhancing the development of wiki-based digital humanities projects (WDHPs), specifically, the projects that focus on gathering contextual information in the culture heritage domain. Design/methodology/approach: An exploratory study was conducted by three steps. A representative entry’s editorial records were reorganized to obtain a data set of discussion statements (n=608), based on which linked-structures were built, and PageRank algorithm was used to analyze the co-editing process. Skewness statistic was applied to measure the consensus of co-editing, and consensus evolution over time was explored. Linear or curve fitting was performed to analyze the correlation between consensus evolution and its influential factors. Findings: In WDHPs, co-editing activity of content creation can be considered as a large-scale group discussion, consensus can evaluate the efficiency of co-editing, which evolves with time and is influenced by the number of statements, breadth and depth of argumentation structure. Taking “Mogao Grottoes” as an example, group discussions around 15 key issues dominate the content creating process, consensus is on a rise with time, finally reaches a relatively high level, and consensus evolution is more influenced by breadth than by depth of argumentation structure, which indicates that co-editing efficiency of “Mogao Grottoes” is fine and more argumentation in a depth manner should be guided. Practical implications: For researchers of WDHPs, it is beneficial to apply online encyclopedia platform combining with consensus analysis to develop WDHPs. For designers of WDHPs, the elements related to argumentation structure can be absorbed into the design to promote co-editing in an effective manner. For DH researchers, the analytic procedure can be beneficial of revealing the interest of contributors in a specific DH field. Originality/value: This research is novel in comprehensively understanding co-editing mechanism of content creation in WDHPs, resulting in a three-step analytic procedure of presenting co-editing process, evaluating and improving co-editing efficiency. © 2019, Emerald Publishing Limited.
语种:
英文
展开