A5037968142- COVID-19 epidemiological studies
- Data-Driven Disease Surveillance
- COVID-19 diagnosis using AI
- Methane Hydrates and Related Phenomena
- Arctic and Antarctic ice dynamics
- Anomaly Detection Techniques and Applications
- Underwater Acoustics Research
- Geographic Information Systems Studies
- Data Management and Algorithms
- Cryospheric studies and observations
- Diverse Scientific Research in Ukraine
- Environmental Monitoring and Data Management
- Environmental Sustainability and Technology
- Air Quality and Health Impacts
- Misinformation and Its Impacts
- Health disparities and outcomes
- Climate Change and Health Impacts
- Regional Socio-Economic Development Trends
- Spatial and Panel Data Analysis
- Advanced Database Systems and Queries
- Influenza Virus Research Studies
- Climate change and permafrost
- Data Analysis with R
University of Colorado Boulder
2020-2024
University of Colorado System
2021
University of California, Santa Barbara
2016-2020
Prediction of the COVID-19 incidence rate is a matter global importance, particularly in United States. As 4 June 2020, more than 1.8 million confirmed cases and over 108 thousand deaths have been reported this country. Few studies examined nationwide modeling States using machine-learning algorithms. Thus, we collected prepared database 57 candidate explanatory variables to examine performance multilayer perceptron (MLP) neural network predicting cumulative rates across continental Our...
Measurements of human interaction through proxies such as social connectedness or movement patterns have proved useful for predictive modeling COVID-19, which is a challenging task, especially at high spatial resolutions. In this study, we develop Spatiotemporal autoregressive model to predict county-level new cases COVID-19 in the coterminous US using spatiotemporal lags infection rates, interactions, mobility, and socioeconomic composition counties features. We capture interactions 1)...
Accurate mapping of sea ice is crucial for marine navigation and monitoring climate change. Automating remains challenging due to remotely-sensed signal ambiguity, the dynamic nature ice, limited field measurements. The AutoICE challenge recently introduced a benchmark advance deep learning mapping. Top-performing solutions used U-Net architecture with extra pre/post-processing steps incorporated location features obtain higher metrics. However, model interpretation diagnostics remain...
Due to the growing volume of remote sensing data and low latency required for safe marine navigation, machine learning (ML) algorithms are being developed accelerate sea ice chart generation, currently a manual interpretation task. However, signal-to-noise ratio freely available Sentinel-1 Synthetic Aperture Radar (SAR) imagery, ambiguity backscatter signals types, scarcity open-source high-resolution labelled makes automating mapping challenging. We use Extreme Earth version 2, benchmark...
Sea ice, crucial to the Arctic and Earth's climate, requires consistent monitoring high-resolution mapping.Manual sea ice mapping, however, is time-consuming subjective, prompting need for automated deep learning-based classification approaches.However, training these algorithms challenging because expert-generated charts, commonly used as data, do not map single types but instead polygons with multiple types.Moreover, distribution of various in charts frequently imbalanced, resulting a...
Up-to-date sea ice charts are crucial for safer navigation in ice-infested waters. Recently, Convolutional Neural Network (CNN) models show the potential to accelerate generation of maps large regions. However, results from CNN still need undergo scrutiny as higher metrics performance not always translate adequate outputs. Sea type classes imbalanced, requiring special treatment during training. We evaluate how three different loss functions, some developed imbalanced class problems, affect...
Sea ice, crucial to the Arctic and Earth's climate, requires consistent monitoring high-resolution mapping. Manual sea ice mapping, however, is time-consuming subjective, prompting need for automated deep learning-based classification approaches. However, training these algorithms challenging because expert-generated charts, commonly used as data, do not map single types but instead polygons with multiple types. Moreover, distribution of various in charts frequently imbalanced, resulting a...
Abstract Measurements of human interaction through proxies such as social connectedness or movement patterns have proved useful for predictive modeling COVID-19. In this study, we first compare the power Facebook’s with cell phone-derived mobility predicting county-level new cases Our experiments show that is a better proxy measuring interactions leading to infections. Next, develop SpatioTemporal autoregressive eXtreme Gradient Boosting (STXGB) model predict COVID-19 in coterminous US. We...
With COVID-19 affecting every country globally and changing everyday life, the ability to forecast spread of disease is more important than any previous epidemic. The conventional methods disease-spread modeling, compartmental models, are based on assumption spatiotemporal homogeneity virus, which may cause forecasting underperform, especially at high spatial resolutions. In this paper we approach task with an alternative technique - machine learning. We present COVID-LSTM, a data-driven...