专用集成电路与系统国家重点实验室
讲座信息：Automation Process of Machine Learning Methods and Big Data Treatment with its Application to Multi-Type Wetland Identification

时 间：2018年5月28日15:15
地 点：张江校区微电子楼389室
演讲人：Professor Sheng-Guo Wang

ABSTRACT
This speech presents his recently published research work with big data and machine learning methods for modeling, prediction and identification of multi-type wetland. Machine learning and Artificial intelligence are discussed. Wetlands are very important to our human beings and ecosystems. Different wetland types play important roles in transportation planning, ecological studies, and conservation planning, which have significant impacts on ecological functions.  
In order to facilitate this identification process, three machine learning methods of Linear Regression Model (LRM, a General Logit), Gradient Boosted Machine (GBM, Stochastic Gradient Boosting) and Random Forest (RF) are studied for the multi-type wetland identification. Taking advantage of high-resolution LiDAR data, which are big, we emphasize topographical information, and develop automation tools for this multi-type wetland identification automation.    The methods and automation tools are applied to a specific study area in North Carolina for validation. We further compare these methods in terms of classification accuracy to provide insights for method selection. The results show that both RF and GBM methods perform well and provide similar results, and the overall accuracies are both above 95%.    The research shows and exemplifies how innovative technologies can be used in lieu of extensive field wetland type delineations and ultimately reduce transportation project delivery time and costs while protecting the environment.  The resulting approach and presented machine learning methods can be applied to broad areas of science, engineering and medicine.

Bio of Professor Sheng-Guo Wang 
Prof. Sheng-Guo Wang is a tenured Full Professor at the University of North Carolina at Charlotte, USA.    Recently, he was a US Fulbright Senior Scholar at The Hong Kong Polytechnic University.    He received his B.S. and M.Sc. in Electrical Engineering from University of Science and Technology of China in 1967 and 1981 respectively, and Ph.D. in Electrical and Computer Engineering from University of Houston, USA in 1994.
Prof. Wang is a recipient of China National Science Conference Award 1978, one of the highest academic honors in China, US Fulbright Senior Scholar Award of 2016-2017, British Council Scholar Award of 1989-1990, and many other academic awards, e.g., his recent NCDOT research project, titled “Improvements to NCDOT’s Wetland Prediction Model” (2012 ~ 2014), has won a US national 2015 “Sweet 16” High Value Research Award recognized by AASHTO (American Association of State Highway and Transportation Officials) and RAC (Research Advisory Committee) in 2015, and also acknowledged at 2016 Transportation Research Board 95th Annual Meeting, National Academies of Sciences-Engineering-Medicine in 2016. 
He has been the Principal Investigators for numerous research projects since 1974.    He has published more than 100 research papers, and his research has been supported by the Fulbright Program, NSF, NCDOT, HP, Tellabs, Agilent, NASA, etc. in US, the British Council in UK, and the China Railways in China.