Spatiotemporal Modeling of the Association between Neighborhood Factors and COVID-19 Incidence Rates in Scotland

This study aims to investigate the association between neighborhood-level factors and COVID-19 incidence in Scotland from a spatiotemporal perspective. The outcome variable is the COVID-19 incidence in Scotland. Based on the identification of the wave peaks for COVID-19 cases between 2020 and 2021, confirmed COVID-19 cases in Scotland can be divided into four phases. To model the COVID-19 incidence, sixteen neighborhood factors are chosen as the predictors. Geographical random forest models are used to examine spatiotemporal variation in major determinants of COVID-19 incidence. The spatial analysis indicates that proportion of religious people is the most strongly associated with COVID-19 incidence in southern Scotland, whereas particulate matter is the most strongly associated with COVID-19 incidence in northern Scotland. Also, crowded households, prepandemic emergency admission rates, and health and social workers are the most strongly associated with COVID-19 incidence in eastern and central Scotland, respectively. A possible explanation is that the association between predictors and COVID-19 incidence might be influenced by local context (e.g., people's lifestyles), which is spatially variant across Scotland. The temporal analysis indicates that dominant factors associated with COVID-19 incidence also vary across different phases, suggesting that pandemic-related policy should take spatiotemporal variations into account. © 2023 by American Association of Geographers.

Earth's magnetic field spatial-temporal changes in relation to the SARS-CoV-2 virus spreading

The results of a study on the possible connection between the spread of the SARS-CoV-2 virus and the Earth's magnetic field based on the analysis of a large array digital data for 95 countries of the world are presented. The dependence of the spatial SARS-CoV-2 virus spread on the magnitude of the BIGRF Earth's main magnetic field modular induction values was established. The maximum diseases number occurs in countries that are located in regions with reduced (25. 0-30. 0 μT) and increased (48. 0-55. 0 μT) values, with a higher correlation for the first case. The spatial dependence of the SARS-CoV-2 virus spreading on geomagnetic field dynamics over the past 70 years was revealed. The maximum diseases number refers to the areas with maximum changes in it, both in decrease direction (up to - 6500 nT) and increase (up to 2500 nT), with a more significant correlation for countries located in regions with increased geomagnetic field. © 2022 EAGE. All Rights Reserved.

Spatio-Temporal Evolution and Driving Factors of "Non-Grain Production” in Hubei Province Based on a Non-Grain Index

Non-grain production (NGP) on cultivated land has become a common phenomenon due to the prosperity of the rural economy and the optimisation of the agricultural structure. However, the excessive use of cultivating land for NGP has threatened food production and the sustainable use of cultivated land. To halt this trend and to ensure food security, the authors of this paper applied a novel non-grain index to measure NGP, which could reflect multiple NGP activities;designated Hubei Province as its object of research;and revealed NGP's spatio-temporal patterns of the past 30 years. We then assessed the characteristics of NGP based on spatial autocorrelation analysis, the Theil index, and geographically weighted regression. The results showed that the value of the non-grain index grew from 0.497 to 1.113 as NGP increased significantly in Hubei Province. The number of high-NGP counties increased, spatial agglomeration became obvious, and the eastern and western sides of Hubei Province witnessed an observable growth in NGP. As a result, the NGP in the eastern and western regions overtook production in the central region. Despite a series of historical subsidy policies and agricultural modernisation initiatives that promoted the planting of grain crops, the policy of "grain on valuable cultivated land” could be better implemented. We conclude by making some suggestions for reducing NGP and protecting cultivated land.

Enhanced cluster detection and noise reduction for geospatial time series data of COVID-19.

Spatial-temporal analysis of the COVID-19 cases is critical to find its transmitting behaviour and to detect the possible emerging clusters. Poisson's prospective space-time analysis has been successfully implemented for cluster detection of geospatial time series data. However, its accuracy, number of clusters, and processing time are still a major problem for detecting small-sized clusters. The aim of this research is to improve the accuracy of cluster detection of COVID-19 at the county level in the U.S.A. by detecting small-sized clusters and reducing the noisy data. The proposed system consists of the Poisson prospective space-time analysis along with Enhanced cluster detection and noise reduction algorithm (ECDeNR) to improve the number of clusters and decrease the processing time. The results of accuracy, processing time, number of clusters, and relative risk are obtained by using different COVID-19 datasets in SaTScan. The proposed system increases the average number of clusters by 7 and the average relative risk by 9.19. Also, it provides a cluster detection accuracy of 91.35% against the current accuracy of 83.32%. It also gives a processing time of 5.69 minutes against the current processing time of 7.36 minutes on average. The proposed system focuses on improving the accuracy, number of clusters, and relative risk and reducing the processing time of the cluster detection by using ECDeNR algorithm. This study solves the issues of detecting the small-sized clusters at the early stage and enhances the overall cluster detection accuracy while decreasing the processing time.

Redes Sociales Geolocalizadas Y Covid-19: Analisis De La Actividad Espaciotemporal De Los Usuarios De Twitter De Espana Durante La Pandemia

This work seeks to show Twitter as an alternative data source for the study of the pandemic caused by the COVID-19 virus in Spain. For this work, an analysis of the spatial and temporal distribution of the sample of users obtained in three different periods of the year 2020 is proposed, and then the obtained results are compared with the same periods of the year prior to the pandemic. A space-time analysis of the use of terms associated with the disease is also elaborated, and heat maps are made to observe the impact caused in the activity of two cities of relevant tourist weight. The obtained results indicate a sharp decrease in the number of users who publish geolocated tweets in the country throughout 2020, especially in the second half of the year and in the interior provinces of the peninsula. A less pronounced decrease in the number of users is also observed in coastal areas and provinces oriented to the tourism sector.

A new approach to COVID-19 data mining: A deep spatial-temporal prediction model based on tree structure for traffic revitalization index.

The outbreak of the COVID-19 epidemic has had a huge impact on a global scale and its impact has covered almost all human industries. The Chinese government enacted a series of policies to restrict the transportation industry in order to slow the spread of the COVID-19 virus in early 2020. With the gradual control of the COVID-19 epidemic and the reduction of confirmed cases, the Chinese transportation industry has gradually recovered. The traffic revitalization index is the main indicator for evaluating the degree of recovery of the urban transportation industry after being affected by the COVID-19 epidemic. The prediction research of traffic revitalization index can help the relevant government departments to know the state of urban traffic from the macro level and formulate relevant policies. Therefore, this study proposes a deep spatial-temporal prediction model based on tree structure for the traffic revitalization index. The model mainly includes spatial convolution module, temporal convolution module and matrix data fusion module. The spatial convolution module builds a tree convolution process based on the tree structure that can contain directional features and hierarchical features of urban nodes. The temporal convolution module constructs a deep network for capturing temporal dependent features of the data in the multi-layer residual structure. The matrix data fusion module can perform multi-scale fusion of COVID-19 epidemic data and traffic revitalization index data to further improve the prediction effect of the model. In this study, experimental comparisons between our model and multiple baseline models are conducted on real datasets. The experimental results show that our model has an average improvement of 21%, 18%, and 23% in MAE, RMSE and MAPE indicators, respectively.

Spatial-Temporal Evolution Patterns and Influencing Factors of Hotels in Yellow River Basin from 2012 to 2022

Governmental attention towards the high-quality development of the Yellow River basin has brought new development opportunities for the hotel industry. This study aims to reveal the spatial-temporal evolution patterns and influencing factors of hotels in the Yellow River Basin from 2012 to 2022, based on economic, social, and physical geographic data of 190,000 hotels in the Yellow River flowing. With the help of a GIS technology system, the spatial-temporal evolution patterns of all hotels, star hotels, and ordinary hotels were explored, respectively. Then, the significant influencing factors of these patterns were revealed by using geographic detector and Person correlation analysis. The following conclusions were drawn: (1) the overall scale of the hotel industry in the Yellow River Basin expanded year by year, achieving rapid growth from 2016, and fluctuating around 2020 due to the impact of the novel coronavirus epidemic;the overall spatial distribution had significant regional differences, showing the structural characteristics of "southeast more, northwest less”;(2) there was a great difference in the degree of spatial autocorrelation agglomeration among prefecture-level cities, and the degree of agglomeration of both the hotel industry as a whole and general hotels decreased year by year, showing a random distribution in 2022;star hotels were always distributed randomly. Additionally, a strong synergistic correlation was shown between the number of ordinary hotels and the number of star hotels in local space;(3) overall, the development of the hotel industry was significantly affected by seven factors: structural force, macro force, ecological force, internal power, consumption power, intermediary power, and external power. There were differences in the forces acting on different types of hotels, which gives a pattern recognition in-depth.

Hourly forecasting of traffic flow rates using spatial temporal graph neural networks.

Traffic congestion forms a large problem in many major metropolitan regions around the world, leading to delays and societal costs. As people resume travel upon relaxation of COVID-19 restrictions and personal mobility returns to levels prior to the pandemic, policy makers need tools to understand new patterns in the daily transportation system. In this paper we use a Spatial Temporal Graph Neural Network (STGNN) to train data collected by 34 traffic sensors around Amsterdam, in order to forecast traffic flow rates on an hourly aggregation level for a quarter. Our results show that STGNN did not outperform a baseline seasonal naive model overall, however for sensors that are located closer to each other in the road network, the STGNN model did indeed perform better.

Spatio-temporal Graph Learning for Epidemic Prediction

The COVID-19 pandemic has posed great challenges to public health services, government agencies, and policymakers, raising huge social conflicts between public health and economic resilience. Policies such as reopening or closure of business activities are formulated based on scientific projections of infection risks obtained from infection dynamics models. Though most parameters in epidemic prediction service models can be set with domain knowledge of COVID-19, a key parameter, namely, human mobility, is often challenging to estimate due to complex spatio-temporal correlations and social contexts under escalating COVID-19 facilities. Moreover, how to integrate the various implicit features to accurately predict infectious cases is still an open issue. To address this challenge, we formulate the problem as a spatio-temporal network representation problem and propose STEP, a Spatio-Temporal Epidemic Prediction framework, to estimate pandemic infection risk of a city by integrating various real-world conditions (e.g., City Risk Index, climate, and medical conditions) into graph-structured data. We also employ a multi-head attention mechanism in representation learning to extract implicit features for a given city. Extensive experiments have been conducted upon the real-world dataset for 51 states (50 states and Washington, D.C.) of the USA. Experimental results show that STEP can yield more accurate pandemic infection risk estimation than baseline methods. Moreover, STEP outperforms other methods in both short-term and long-term prediction. [ABSTRACT FROM AUTHOR] Copyright of ACM Transactions on Intelligent Systems & Technology is the property of Association for Computing Machinery and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

COVID-19 Data Analysis Using Bayesian Models and Nonparametric Geostatistical Models

Since December 2019, many statistical spatial–temporal methods have been developed to track and predict the spread of the COVID-19 pandemic. In this paper, we analyzed the COVID-19 dataset which includes the number of biweekly infected cases registered in Ontario from March 2020 to the end of June 2021. We made use of Bayesian Spatial–temporal models and Area-to-point (ATP) and Area-to-area (ATA) Poisson Kriging models. With the Bayesian models, spatial–temporal effects and government intervention effects on infection risk are considered while the ATP Poisson Kriging models are used to display the spread of the pandemic over space. © 2023 by the authors.

[Impacts of Emission and Meteorological Conditions on Air Pollutants at Various Sites Around the COVID-19 Lockdown in Wuhan].

The random forest algorithm was used to separate the mass concentrations of six air pollutants (SO2, NO2, CO, PM10, PM2.5, and O3) contributed by emissions and meteorological conditions. Their variations for five types of sites including Wuhan's central urban, suburb, industrial, the third ring road traffic, and urban background sites were investigated. The results showed that the values of PM2.5/CO, PM10/CO, and NO2/CO during the lockdown period decreased by 10.8-21.7, 9.34-24.7, and 14.4-22.1 times compared with the period before the lockdown, indicating that the contributions of emissions to PM2.5, PM10, and NO2 were reduced. O3/CO increased by 50.1-61.5 times, implying that the secondary formation increased obviously. The contributions of emissions to various types of pollutants all increased after the lockdown. During the lockdown period, affected by the operation of some uninterrupted industrial processes, PM2.5 concentrations in industrial areas dropped the least (20.5%). Compared with the lockdown period, residential activities, transportation, and industrial production were basically restored after the lockdown, resulting in the alleviation of the reduction in PM2.5 emission-related concentrations. The increase in emission-related O3 concentrations could be associated with the decreased NO and PM2.5 concentrations during the lockdown period. The elevated O3 partially offset the improved air quality brought by the reduced NO2and PM2.5 concentrations. After the lockdown, ρ(O3) related with meteorology at the suburban and urban background sites increased by 16.2 µg·m-3 and 16.1 µg·m-3, respectively, which could be attributed to the increased ambient temperature and decreased relative humidity. The decrease in PM2.5 and increase in O3 concentrations caused by reduced traffic and industrial emissions at the third ring road traffic and central urban regions can provide reference for the current coordinated and precise control of PM2.5 and O3 in subregions.

Spatial and temporal distribution of emerging airborne viral infectious diseases outbreaks on a global scale.

Aim: This study aimed to explore the spatial and temporal characteristics of emerging airborne viral infectious diseases outbreaks worldwide. Subject and methods: We conducted a systematic literature review on outbreaks of emerging airborne viral infectious diseases and calculated outbreak number and intensity at the country level. Fisher's exact test was used to compare the viral infectious diseases outbreaks in different income-level regions. To identify the major airborne viral infectious diseases outbreaks, we ranked and extracted the leading viral infectious diseases in outbreak number and intensity in each country by year. Results: A total of 2505 outbreaks were reported from 1873 to 2021 across 2010 studies. There were 47 countries (47/130, 36.15%) with more frequent emerging airborne viral infectious disease outbreaks (more than nine outbreaks), and these countries mainly distributed in high-income regions (22/47 countries, 46.81%, p < 0.05), especially in Western Europe (14/47 countries, 29.79%, p < 0.05). The number of overall outbreaks was more in the United States and China than in other countries in different years. Outbreaks of measles and influenza are always frequent and intense. Highly pathogenic human coronaviruses infection caused short-term pandemics during which their outbreak number and intensity exceeded other viruses. Rift valley fever outbreaks in the human population are spreading outside of Africa through the flow of goods and travelers. Conclusion: Countries in high-income regions reported more emerging airborne viral infectious diseases outbreaks, especially in the Western European region, the United States, and China. It is urgent to strengthen collaborative surveillance of emerging airborne viruses, cross-border flow of goods and travelers, and ecological environment to avoid the spread of viral infectious diseases outbreaks worldwide. Supplementary Information: The online version contains supplementary material available at 10.1007/s10389-023-01850-3.

Optimizing Large-Scale COVID-19 Nucleic Acid Testing with a Dynamic Testing Site Deployment Strategy.

The COVID-19 epidemic has spread worldwide, infected more than 0.6 billion people, and led to about 6 million deaths. Conducting large-scale COVID-19 nucleic acid testing is an effective measure to cut off the transmission chain of the COVID-19 epidemic, but it calls for deploying numerous nucleic acid testing sites effectively. In this study, we aim to optimize the large-scale nucleic acid testing with a dynamic testing site deployment strategy, and we propose a multiperiod location-allocation model, which explicitly considers the spatial-temporal distribution of the testing population and the time-varied availability of various testing resources. Several comparison models, which implement static site deployment strategies, are also developed to show the benefits of our proposed model. The effectiveness and benefits of our model are verified with a real-world case study on the Chenghua district of Chengdu, China, which indicates that the optimal total cost of the dynamic site deployment strategy can be 15% less than that of a real plan implemented in practice and about 2% less than those of the other comparison strategies. Moreover, we conduct sensitivity analysis to obtain managerial insights and suggestions for better testing site deployment in field practices. This study highlights the importance of dynamically deploying testing sites based on the target population's spatial-temporal distribution, which can help reduce the testing cost and increase the robustness of producing feasible plans with limited medical resources.

Investigation on the spatial and temporal patterns of coupling sustainable development posture and economic development in World Natural Heritage Sites: A case study of Jiuzhaigou, China

To continue directing global sustainable development efforts from 2015 to 2030, the United Nations adopted 17 global development goals known as the Sustainable Development Goals (SDGs) when the Millennium Development Goals (MDGs) from 2000 to 2015 expired. Sustainable development of World Natural Heritage Sites is one of these 17 MDGs and a crucial step toward achieving global sustainability. A scientific and systematic indicator system that can measure the sustainable development of natural World Heritage Sites more objectively and fairly is urgently needed to support the establishment of SDG11.4 on a Chinese scale and to help with the subsequent promotion of the development of natural World Heritage Sites. This study proposes a comprehensive assessment indicator system for the sustainable development of natural heritage sites based on the theoretical framework of "value contribution-environmental effect” to quantify the sustainable development of natural heritage sites. The study is based on the ecological environment and regional economic and social data of Jiuzhaigou World Natural Heritage Site from 2010 to 2020. Finally, the degree of coupling and coordination between the natural environment and economic development is assessed and studied. The results show that tourism to the World Heritage Site drove rapid economic development in Jiuzhaigou County between 2010 and 2020. As the fame of the World Heritage Site Jiuzhaigou has grown, so has the per capita income of local locals, making them unduly reliant on tourists for a living. Meanwhile, both the 2017 earthquake and the COVID-19 epidemic in 2019 have had substantial detrimental effects on the local economy. Furthermore, the Jiuzhaigou sustainable development trend from 2010 to 2020 exhibits a "W-shaped” curve, and there is a high level of positive coupling between the Jiuzhaigou sustainable development trend and economic development, and the two are mutually reinforcing.

Spatial-temporal clustering of an outbreak of SARS-CoV-2 Delta VOC in Guangzhou, China in 2021.

Background: In May 2021, the SARS-CoV-2 Delta variant led to the first local outbreak in China in Guangzhou City. We explored the epidemiological characteristics and spatial-temporal clustering of this outbreak. Methods: Based on the 153 cases in the SARS-CoV-2 Delta variant outbreak, the Knox test was used to analyze the spatial-temporal clustering of the outbreak. We further explored the spatial-temporal clustering by gender and age groups, as well as compared the changes of clustering strength (S) value between the two outbreaks in Guangzhou. Results: The result of the Knox analysis showed that the areas at short distances and brief periods presented a relatively high risk. The strength of clustering of male-male pairs was higher. Age groups showed that clustering was concentrated in cases aged ≤ 18 years matched to 18-59 years and cases aged 60+ years. The strength of clustering of the outbreak declined after the implementation of public health measures. The change of strength of clustering at time intervals of 1-5 days decreased greater in 2021 (S = 129.19, change rate 38.87%) than that in 2020 (S = 83.81, change rate 30.02%). Conclusions: The outbreak of SARS-CoV-2 Delta VOC in Guangzhou has obvious spatial-temporal clustering. The timely intervention measures are essential role to contain this outbreak of high transmission.

A Study on the Spatial-Temporal Evolution and Driving Factors of Non-Grain Production in China's Major Grain-Producing Provinces.

Food self-sufficiency in a large country with 1.4 billion people is very important for the Chinese government, especially in the context of COVID-19 and the Russian-Ukrainian conflict. The objective of this paper is to explore the spatial-temporal evolution and driving factors of non-grain production in thirteen major grain-producing provinces in China, which account for more than 75% of China's grain production, using 2011-2020 prefecture-level statistics. In the present study, the research methodology included GIS spatial analysis, hot spot analysis, and spatial Durbin model (SDM). The findings of this study are as follows: (1) The regions with a higher level of non-grain production were mainly concentrated in the central and western regions of Inner Mongolia, the middle and lower reaches of Yangtze River and Sichuan, while the regions with a low level of non-grain production were mainly distributed in the Northeast Plain. The regions with a higher proportion of grain production to the national total grain production were concentrated in the Northeast Plain, the North China Plain, and the Middle and Lower Yangtze River Plain of China. The hot spot regions with changes in non-grain production levels were mainly distributed in the Sichuan region and Alashan League City in Inner Mongolia, and the cold spot regions were mainly distributed in Hebei, Shandong, Henan, and other regions. (2) An analysis of the SDM indicated that the average air temperature among the natural environment factors, the ratio of the sum of gross secondary and tertiary industries to GDP, the ratio of gross primary industry to the GDP of economic development level, the urbanization rate of social development, and the difference in disposable income per capita between urban and rural residents of the urban-rural gap showed positive spatial spillover effects. The grain yield per unit of grain crop sown area of grain production resource endowment, the total population of social development, and the area sown to grain crops per capita of grain production resource endowment all showed negative spatial spillover effects. The research results of this paper can provide a reference for the country to carry out the governance of non-grain production and provide a reference for China's food security guarantee.

COVID-19 SPATIOTEMPORAL SIR MODEL: REGIONAL OPTIMAL CONTROL APPROACH

In the present work, we consider a spatio-temporal model to describe the evolution of covid19 in an area Ω (Ω can be a city, a country,..). Taking into account the financial means of the considered country, we suppose that the number of available vaccines is destined to a region ω1 ⊆Ω (ω1 can be an industrial city, a university city. ..) and we suppose that the available treatments are dedicated to a region ω2 ⊆ Ω (ω2 can be a military city,..), it is not excluded that ω1 = ω2 . To minimize the number of infection with minimal cost, we apply an optimal regional control strategy to stop the death of infected individuals in the considered area. Much of this work has been devoted to mathematical study, where the existence of the optimal controls and the solutions of the state system are proven, an optimal control characterization in terms of state and adjoint functions are provided, and the optimality system is solved numerically using a forward-backward sweep method. Our numerical results suggest that when vaccination and treatment procedures are used together, the control approach becomes more effective in protecting a specific region from epidemic transmission from neighboring regions. © 2022 the author(s).

Predicting COVID-19 community infection relative risk with a Dynamic Bayesian Network.

As COVID-19 continues to impact the United States and the world at large it is becoming increasingly necessary to develop methods which predict local scale spread of the disease. This is especially important as newer variants of the virus are likely to emerge and threaten community spread. We develop a Dynamic Bayesian Network (DBN) to predict community-level relative risk of COVID-19 infection at the census tract scale in the U.S. state of Indiana. The model incorporates measures of social and environmental vulnerability-including environmental determinants of COVID-19 infection-into a spatial temporal prediction of infection relative risk 1-month into the future. The DBN significantly outperforms five other modeling techniques used for comparison and which are typically applied in spatial epidemiological applications. The logic behind the DBN also makes it very well-suited for spatial-temporal prediction and for "what-if" analysis. The research results also highlight the need for further research using DBN-type approaches that incorporate methods of artificial intelligence into modeling dynamic processes, especially prominent within spatial epidemiologic applications.

Monitoring gas flaring in Texas using time-series sentinel-2 MSI and landsat-8 OLI images

Since the shale Oil/Gas revolution, gas flaring and venting in the United States has garnered increasing attention. There is a pressing need to understand the spatial–temporal characteristics of gas flaring and track the associated greenhouse gas emissions. In this context, we use a thermal anomaly index (TAI) incorporating the Google Earth Engine (GEE) cloud computation and local batch processing for monitoring gas flaring and characterizing its spatial–temporal dynamics. We then apply a quantitative analysis of satellite-based carbon dioxide (CO2) and methane (CH4) in the gas flaring region. Here, we generate a gas flaring sites inventory in Texas from 2013 to 2022 based on > 83,500 multi-source moderate-resolution images (including 74,627 Sentinel-2 Multispectral Instrument [MSI] images and 8,969 Landsat-8 Operational Land Imager [OLI] images). Validations and comparisons demonstrate that our method is reliable for MSI and OLI images, with an overall accuracy of > 95 % and a low commission rate and omission rate. We detected 217,034 gas flares from 9,296 flaring sites in Texas, and the majority (>92 %) were found in the central and western regions of the Permian Basin and the Eagle Ford Shale. The number of detected gas flaring sites vastly outnumbered the existing Visible Infrared Imaging Radiometer Suite (VIIRS) fire products, with an upward trend from 2013 to 2019 and a downward trend from 2020 to 2022. Notably, the gas flaring sites dropped significantly at the beginning of the COVID-19 pandemic (from December 2019 to May 2020), with the lowest average monthly growth rate of −14.38 %, and fell to the level of mid-2017. Application of gas flaring data identifies the localized greenhouse gas (GHG) emission hotspots in Texas and demonstrates that the increased effect of CH4 released from gas flaring regions was significantly stronger than that of CO2. These findings can provide references for monitoring similar small industrial sources in the future, can be used as an essential supplement to low-resolution fire products, and improve our understanding of CO2 and CH4 emissions from gas flaring at fine spatial scales.

Spatial-Temporal Synchronous Graph Transformer network (STSGT) for COVID-19 forecasting.

COVID-19 has become a matter of serious concern over the last few years. It has adversely affected numerous people around the globe and has led to the loss of billions of dollars of business capital. In this paper, we propose a novel Spatial-Temporal Synchronous Graph Transformer network (STSGT) to capture the complex spatial and temporal dependency of the COVID-19 time series data and forecast the future status of an evolving pandemic. The layers of STSGT combine the graph convolution network (GCN) with the self-attention mechanism of transformers on a synchronous spatial-temporal graph to capture the dynamically changing pattern of the COVID time series. The spatial-temporal synchronous graph simultaneously captures the spatial and temporal dependencies between the vertices of the graph at a given and subsequent time-steps, which helps capture the heterogeneity in the time series and improve the forecasting accuracy. Our extensive experiments on two publicly available real-world COVID-19 time series datasets demonstrate that STSGT significantly outperforms state-of-the-art algorithms that were designed for spatial-temporal forecasting tasks. Specifically, on average over a 12-day horizon, we observe a potential improvement of 12.19% and 3.42% in Mean Absolute Error (MAE) over the next best algorithm while forecasting the daily infected and death cases respectively for the 50 states of US and Washington, D.C. Additionally, STSGT also outperformed others when forecasting the daily infected cases at the state level, e.g., for all the counties in the State of Michigan. The code and models are publicly available at https://github.com/soumbane/STSGT.