School of Science Department of Mathematics 44 Estimation of Wet Deposition Chemical Components in Northern Hemisphere by Using Deep-learning Technique Supervisor: FUNG Jimmy Chi Hung / MATH Student: PANG Ran / MAEC Course: UROP 1100, Spring PM2.5 (PM mass less than 2.5 μm in aerodynamic diameter) has a detrimental effect on the various environmental components, posing significant concerns. In China, there has been a rise in PM2.5 level in 2023. In this progress report, an ensemble back-propagation neural network (eBPNN) was reproduced to estimate PM2.5 concentration over China and get a monthly prediction using multi-source data from January to February 2015. The correlation between AOD (aerosol optical depth), meteorological data, and PM2.5 is obtained. The R2 value for the 10-fold cross-validation of PM2.5 is 0.6715. Results indicate a higher concentration of PM2.5 in the eastern China. Therefore, the area concerned should strengthen its air quality management. Improvement of the Air Quality Forecast by Using Deep-learning Technique Supervisor: FUNG Jimmy Chi Hung / MATH Student: LI Jiaran / DSCT Course: UROP 1100, Summer In this UROP, a brief idea about using deep learning techniques to forecast pm2.5 value is introduced. Since the measurement of pm2.5 value is difficult in some location, but it is feasible to measure other variables such as temperature, pressure, etc. It is good for us to develop a model to predict pm2.5 using these easier measured parameters. Improvement of the Air Quality Forecast by Using Deep-learning Technique Supervisor: FUNG Jimmy Chi Hung / MATH Student: SHI Boqing / QSA Course: UROP 1100, Spring In this semester, some foundation of required skillset and machine learning algorithm for segmentation were learnt and tested out. We tried out the image segmentation algorithm from Keras and we got ourselves familiar with using GPU for machine learning and understand script and the theory behind image segmentation. In the coming semester, the script would be modified so that it can be used for real building data and apply to the Hong Kong region.
RkJQdWJsaXNoZXIy NDk5Njg=