School of Science Department of Mathematics 43 A Machine Learning Approach to Study the Relationship Between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: WANG Yuanfu / MAEC Course: UROP 3200, Fall UROP 4100, Spring Remote sensing using satellite images can be used to characterize urban morphology when there is no survey data available. Focusing on Hong Kong, this study proposes a machine learning approach to retrieving a 3dimensional urban morphology dataset with a 100-meter resolution by predicting the area-weighted building height. The Sentinel-1A satellite images can capture the temporal change of shadowed areas cast by buildings, implying the building height in the surrounding areas. For accuracy, different machine-learning algorithms are proposed to be used in modeling the building heights. A Machine Learning Approach to Study the Relationship Between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: XIAO Xiuquan / MATH-STAT Course: UROP 1100, Fall UROP 2100, Spring This study utilizes a Machine Learning approach to investigate the urban heat island (UHI) effect in the Pearl River Delta region, using high-resolution land surface temperature data and urban morphological parameters. A Random Forest model is developed to predict heat distributions, integrating environmental and geographical data to enhance accuracy. The model demonstrates improved predictive capabilities but exhibits limitations in generalizing to new urban areas. Future research directions involve refining data interpolation techniques and adjusting model parameters to further improve accuracy and applicability in diverse urban settings. Estimation of Wet Deposition Chemical Components in Northern Hemisphere by Using Deep-learning Technique Supervisor: FUNG Jimmy Chi Hung / MATH Student: CHANG Joshua / QFIN Course: UROP 1100, Summer In this UROP project, I built and trained a simple linear neural network to estimate the amount of PM2.5 particles present, given different meteorological data. After cleaning up the data, building and training the model, the resulting mean-squared error loss was 2791.41.
RkJQdWJsaXNoZXIy NDk5Njg=