babyÖ±²¥app

My name is Zhongying Wang. I am currently a third-year PhD student at the University of babyÖ±²¥app, Boulder, where I have had the privilege of immersing myself in the vibrant academic environment of the Geography department. Through advanced coursework, including specialized classes like spatial machine learning and applied deep learning, I have acquired expertise in cutting-edge techniques crucial for my research. Under the mentorship of Dr. Morteza Karimzadeh at the GeoHAI Lab, I've explored areas such as infectious disease forecasting and air pollution downscaling, leveraging my interdisciplinary background to tackle pressing environmental and public health challenges effectively.

My research over the past two years has been focused on improving COVID-19 hospitalization forecasting, a critical component of pandemic response and preparedness. Building upon previous work in COVID-19 case forecasting, I have led the development of predictive models specifically tailored to anticipate hospitalization rates, aiming to better prepare for  the virus's impact on healthcare systems. Using data-driven methodologies, particularly deep learning techniques, we have designed a novel Long Short-Term Memory (LSTM) framework to forecast daily incident hospitalizations at the state level across the U.S. What sets our work apart is the incorporation of a unique spatiotemporal features, built on data from Facebook's social connectedness dataset. This innovative feature acts as a proxy for population mobility and interaction across state lines, enabling us to effectively capture transmission dynamics across various spatial and temporal scales in our predictive models.

Our involvement in the , a collaborative initiative, was aimed at putting our research into practice. The Centers for Disease Control and Prevention (CDC) used forecasts submitted to the Forecast Hub, serving as a vital resource for public health authorities and policymakers, furnishing them with essential insights for informed decision-making. Importantly, our model's superior performance during the submission period from February 27, 2022, to June 10, 2023 as evaluated by external researchers at the Forecast Hub, highlights the innovative nature of our models and spatial considerations, in enhancing predictive accuracy for complex public health challenges. By leveraging advanced AI techniques and integrating insights from social media data, our models enrich this collaborative effort by providing actionable insights into hospitalization trends, thereby facilitating targeted interventions to mitigate the spread of the virus and alleviate its impact on communities.