报告地点：Tencent会议711 686 197
Personalized Big Data Analytics in Cyber-Social Computing Applications(周晓康)
Machine Learning and Deep Learning based Forecasting with Time Series Data (Ke Yan)
周晓康，现任日本滋贺大学数据科学学院副教授。2014年毕业于日本早稻田大学，获人类信息科学(Human Sciences)博士学位。2012至2015年，于早稻田大学人间科学学术院任研究助手(Research Associate)。2017年起，于日本理化研究所革新知能综合研究中心（AIP）兼职任客员研究员。主要从事计算机科学，数据科学，及社会人类信息学的跨学科多领域研究工作。研究兴趣包括：大数据、机器学习、行为认知、普适计算智能与安全。发表学术论文100余篇，其中SCI期刊论文50余篇(中科院1区，IEEE/ACM Trans 31篇，高被引5篇，热点3篇)，包括IEEE Transactions on Human-Machine Systems, IEEE Transactions on Learning Technologies, IEEE Transactions on Computational Social Systems, IEEE Transactions on Emerging Topics in Computing, IEEE Internet of Things Journal, IEEE Transactions on Services Computing, IEEE Transactions on Big Data, IEEE Transactions on Industrial Informatics, IEEE/ACM Transactions on Computational Biology and Bioinformatics, IEEE Transactions on Network Science and Engineering, IEEE Transactions on Vehicular Technology, ACM Transactions on Multimedia Computing, Communications, and Applications等。发表论文多次在国际学术会议，如IEEE DependSys’21, EAI CloudComp’21, IEEE SmartData’20, ICADIWT’16, IEEE ITME’14, AIM’13, IET U-Media’12中荣获最佳论文奖，CENet’21中获优秀论文奖。2020年，获IEEE SMC Society Andrew P. Sage Best Transactions Paper Award汇刊最佳论文奖, IEEE TCSC Award for Excellence for Early Career Researchers优秀青年科学家。近年来，在多个国际知名期刊，如FGCS, JPDC, MTAP, Ad Hoc Networks, WWW, CAEE, BDR, INF, IEM, TCBB, BAE等担任客座编委，目前在HCIS, AIHC, CAEE等国际著名期刊担任副编委，并于多个IEEE重要国际学术会议担任程序委员会主席。目前为美国IEEE CS，ACM，日本IPSJ, JSAI，中国CCF会员。
Ke Yan is currently an assistant professor with the department of Building, School of Design and Environment, National University of Singapore (NUS). He is also a visiting professor of the Waseda University, Tokyo, Japan and Huaqiao University, China. Dr. Yan is largely engaged in cross-discipline research fields, including machine learning, artificial intelligence, cyber intelligence, applied mathematics, sustainability and applied energy. He is actively involved with multiple highly-ranked journals’ editorial boards, such as the IEEE Transactions on Industrial Informatics (TII) and IEEE/ACM Transactions on Computational Biology and Bioinformatics. He has published more than 70 full length papers with highly ranked conferences and journals, including Association for the Advancement of Artificial Intelligence (AAAI), IEEE Transactions on Industrial Informatics (TII), IEEE Transactions on Sustainable Energy (TSE), IEEE Transactions on Systems, Man and Cybernetics: Systems (SMCA) and Applied Energy (AE).
The high development of emerging computing paradigms, such as Ubiquitous Computing, Mobile Computing, and Social Computing, has brought us a big change from all walks of our work, life, learning and entertainment. Especially, with the high accessibility of social networking services, more and more populations have been engaged into this kind of integration of real physical world and cyber digital space. In this talk, we concentrate on personalized big data analytics in cyber-social computing applications, specifically, discuss the research on scholarly big data, which is a large-scale collection of academic information, technical data, and collaboration relationships.Mechanisms and algorithms are introduced to facilitate the adoption of social computing paradigm that made it easier for researchers to join collaborative research activities and share academic data across the highly interlaced cyber-social networks.
Artificial intelligence (AI) technologies, including machine learning (ML) and deep learning (DL), bring the great advantage in forecasting to various applications of time series analysis. In the seminar, first, we will introduce the fundamental knowledge of time series data and time series analysis. Traditional methods, such as the regression analysis and moving average model are introduced. Then, real-world time series applications, including energy consumption forecasting, solar energy PV system optimization and air quality forecasting, will be visited in this seminar. Last, the cutting-edge AI techniques, such as the bidirectional long short term memory (BiLSTM) and the nested long short term memory (NLSTM) neural networks, are presented to show the advantages over traditional ML and DL approaches. A wavelet-transform based data decomposition and a multi-tasking neural network structure are designed for better performances in forecasting.