icumonitoring.ch: a platform for short-term forecasting of intensive care unit occupancy during the COVID-19 epidemic in Switzerland.

In Switzerland, the COVID-19 epidemic is progressively slowing down owing to “social distancing” measures introduced by the Federal Council on 16 March 2020. However, the gradual ease of these measures may initiate a second epidemic wave, the length and intensity of which are difficult to anticipate. In this context, hospitals must prepare for a potential increase in intensive care unit (ICU) admissions of patients with acute respiratory distress syndrome. Here, we introduce icumonitoring.ch, a platform providing hospital-level projections for ICU occupancy. We combined current data on the number of beds and ventilators with canton-level projections of COVID-19 cases from two S-E-I-R models. We disaggregated epidemic projection in each hospital in Switzerland for the number of COVID-19 cases, hospitalisations, hospitalisations in ICU, and ventilators in use. The platform is updated every 3-4 days and can incorporate projections from other modelling teams to inform decision makers with a range of epidemic scenarios for future hospital occupancy.

Deep reinforcement learning stock market trading, utilizing a CNN with candlestick images.

Billions of dollars are traded automatically in the stock market every day, including algorithms that use neural networks, but there are still questions regarding how neural networks trade. The black box nature of a neural network gives pause to entrusting it with valuable trading funds. A more recent technique for the study of neural networks, feature map visualizations, yields insight into how a neural network generates an output. Utilizing a Convolutional Neural Network (CNN) with candlestick images as input and feature map visualizations gives a unique opportunity to determine what in the input images is causing the neural network to output a certain action. In this study, a CNN is utilized within a Double Deep Q-Network (DDQN) to outperform the S&P 500 Index returns, and also analyze how the system trades. The DDQN is trained and tested on the 30 largest stocks in the S&P 500. Following training the CNN is used to generate feature map visualizations to determine where the neural network is placing its attention on the candlestick images. Results show that the DDQN is able to yield higher returns than the S&P 500 Index between January 2, 2020 and June 30, 2020. Results also show that the CNN is able to switch its attention from all the candles in a candlestick image to the more recent candles in the image, based on an event such as the coronavirus stock market crash of 2020.

Gauging the Impact of Artificial Intelligence and Mathematical Modeling in Response to the COVID-19 Pandemic: A Systematic Review.

While the world continues to grapple with the devastating effects of the SARS-nCoV-2 virus, different scientific groups, including researchers from different parts of the world, are trying to collaborate to discover solutions to prevent the spread of the COVID-19 virus permanently. Henceforth, the current study envisions the analysis of predictive models that employ machine learning techniques and mathematical modeling to mitigate the spread of COVID-19. A systematic literature review (SLR) has been conducted, wherein a search into different databases, viz., PubMed and IEEE Explore, fetched 1178 records initially. From an initial of 1178 records, only 50 articles were analyzed completely. Around (64%) of the studies employed data-driven mathematical models, whereas only (26%) used machine learning models. Hybrid and ARIMA models constituted about (5%) and (3%) of the selected articles. Various Quality Evaluation Metrics (QEM), including accuracy, precision, specificity, sensitivity, Brier-score, F1-score, RMSE, AUC, and prediction and validation cohort, were used to gauge the effectiveness of the studied models. The study also considered the impact of Pfizer-BioNTech (BNT162b2), AstraZeneca (ChAd0x1), and Moderna (mRNA-1273) on Beta (B.1.1.7) and Delta (B.1.617.2) viral variants and the impact of administering booster doses given the evolution of viral variants of the virus.

Modeling and forecasting the total number of cases and deaths due to pandemic.

The COVID-19 pandemic has appeared as the predominant disease of the 21st century at the end of 2019 and was a drastic start with thousands of casualties and the COVID-19 victims in 2020. Due to the drastic effect, COVID-19 scientists are trying to work on pandemic diseases and Governments are interested in the development of methodologies that will minimize the losses and speed up the process of cure by providing vaccines and treatment for such pandemics. The development of a new vaccine for any pandemic requires long in vitro and in vivo trials to use. Thus the strategies require understanding how the pandemic is spreading in terms of affected cases and casualties occurring from this disease, here we developed a forecasting model that can predict the no of cases and deaths due to pandemic and that can help the researcher, government, and other stakeholders to devise their strategies so that the damages can be minimized. This model can also be used for the judicial distribution of resources as it provides the estimates of the number of casualties and number of deaths with high accuracy, Government and policymakers on the basis of forecasted value can plan in a better way. The model efficiency is discussed on the basis of the available dataset of John Hopkins University repository in the period when the disease was first reported in the six countries till the mid of May 2020, the model was developed on the basis of this data, and then it is tested by forecasting the no of deaths and cases for next 7 days, where the proposed strategy provided excellent forecasting. The forecast models are developed for six countries including Pakistan, India, Afghanistan, Iran, Italy, and China using polynomial regression of degrees 3-5. But the models are analyzed up to the 6th-degree and the suitable models are selected based on higher adjusted R-square (R2 ) and lower root-mean-square error and the mean absolute percentage error (MAPE). The values of R2 are greater than 99% for all countries other than China whereas for China this R2 was 97%. The high values of R2 and Low value of MAPE statistics increase the validity of proposed models to forecast the total no cases and total no of deaths in all countries. Iran, Italy, and Afghanistan also show a mild decreasing trend but the number of cases is far higher than the decrease percentage. Although India is expected to have a consistent result, more or less it depicts some other biasing factors which should be figured out in separate research.

Forecasting emergence of COVID-19 variants of concern.

We consider whether one can forecast the emergence of variants of concern in the SARS-CoV-2 outbreak and similar pandemics. We explore methods of population genetics and identify key relevant principles in both deterministic and stochastic models of spread of infectious disease. Finally, we demonstrate that fitness variation, defined as a trait for which an increase in its value is associated with an increase in net Darwinian fitness if the value of other traits are held constant, is a strong indicator of imminent transition in the viral population.

The research of ARIMA, GM(1,1), and LSTM models for prediction of TB cases in China.

BACKGROUND AND OBJECTIVE: Tuberculosis (Tuberculosis, TB) is a public health problem in China, which not only endangers the population's health but also affects economic and social development. It requires an accurate prediction analysis to help to make policymakers with early warning and provide effective precautionary measures. In this study, ARIMA, GM(1,1), and LSTM models were constructed and compared, respectively. The results showed that the LSTM was the optimal model, which can be achieved satisfactory performance for TB cases predictions in mainland China. METHODS: The data of tuberculosis cases in mainland China were extracted from the National Health Commission of the People's Republic of China website. According to the TB data characteristics and the sample requirements, we created the ARIMA, GM(1,1), and LSTM models, which can make predictions for the prevalence trend of TB. The mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) were applied to evaluate the effects of model fitting predicting accuracy. RESULTS: There were 3,021,995 tuberculosis cases in mainland China from January 2018 to December 2020. And the overall TB cases in mainland China take on a downtrend trend. We established ARIMA, GM(1,1), and LSTM models, respectively. The optimal ARIMA model is the ARIMA (0,1,0) × (0,1,0)12. The equation for GM(1,1) model was X(k+1) = -10057053.55e(-0.01k) + 10153178.55 the Mean square deviation ratio C value was 0.49, and the Small probability of error P was 0.94. LSTM model consists of an input layer, a hidden layer and an output layer, the parameters of epochs, learning rating are 60, 0.01, respectively. The MAE, RMSE, and MAPE values of LSTM model were smaller than that of GM(1,1) and ARIMA models. CONCLUSIONS: Our findings showed that the LSTM model was the optimal model, which has a higher accuracy performance than that of ARIMA and GM (1,1) models. Its prediction results can act as a predictive tool for TB prevention measures in mainland China.

A proficient approach to forecast COVID-19 spread via optimized dynamic machine learning models.

This study aims to develop an assumption-free data-driven model to accurately forecast COVID-19 spread. Towards this end, we firstly employed Bayesian optimization to tune the Gaussian process regression (GPR) hyperparameters to develop an efficient GPR-based model for forecasting the recovered and confirmed COVID-19 cases in two highly impacted countries, India and Brazil. However, machine learning models do not consider the time dependency in the COVID-19 data series. Here, dynamic information has been taken into account to alleviate this limitation by introducing lagged measurements in constructing the investigated machine learning models. Additionally, we assessed the contribution of the incorporated features to the COVID-19 prediction using the Random Forest algorithm. Results reveal that significant improvement can be obtained using the proposed dynamic machine learning models. In addition, the results highlighted the superior performance of the dynamic GPR compared to the other models (i.e., Support vector regression, Boosted trees, Bagged trees, Decision tree, Random Forest, and XGBoost) by achieving an averaged mean absolute percentage error of around 0.1%. Finally, we provided the confidence level of the predicted results based on the dynamic GPR model and showed that the predictions are within the 95% confidence interval. This study presents a promising shallow and simple approach for predicting COVID-19 spread.

Fundamental limitations on efficiently forecasting certain epidemic measures in network models.

The ongoing COVID-19 pandemic underscores the importance of developing reliable forecasts that would allow decision makers to devise appropriate response strategies. Despite much recent research on the topic, epidemic forecasting remains poorly understood. Researchers have attributed the difficulty of forecasting contagion dynamics to a multitude of factors, including complex behavioral responses, uncertainty in data, the stochastic nature of the underlying process, and the high sensitivity of the disease parameters to changes in the environment. We offer a rigorous explanation of the difficulty of short-term forecasting on networked populations using ideas from computational complexity. Specifically, we show that several forecasting problems (e.g., the probability that at least a given number of people will get infected at a given time and the probability that the number of infections will reach a peak at a given time) are computationally intractable. For instance, efficient solvability of such problems would imply that the number of satisfying assignments of an arbitrary Boolean formula in conjunctive normal form can be computed efficiently, violating a widely believed hypothesis in computational complexity. This intractability result holds even under the ideal situation, where all the disease parameters are known and are assumed to be insensitive to changes in the environment. From a computational complexity viewpoint, our results, which show that contagion dynamics become unpredictable for both macroscopic and individual properties, bring out some fundamental difficulties of predicting disease parameters. On the positive side, we develop efficient algorithms or approximation algorithms for restricted versions of forecasting problems.

Forecasting the Severity of COVID-19 Pandemic Amidst the Emerging SARS-CoV-2 Variants: Adoption of ARIMA Model.

Currently, the global report of COVID-19 cases is around 110 million, and more than 2.43 million related death cases as of February 18, 2021. Viruses continuously change through mutation; hence, different virus of SARS-CoV-2 has been reported globally. The United Kingdom (UK), South Africa, Brazil, and Nigeria are the countries from which these emerged variants have been notified and now spreading globally. Therefore, these countries have been selected as a research sample for the present study. The datasets analyzed in this study spanned from March 1, 2020, to January 31, 2021, and were obtained from the World Health Organization website. The study used the Autoregressive Integrated Moving Average (ARIMA) model to forecast coronavirus incidence in the UK, South Africa, Brazil, and Nigeria. ARIMA models with minimum Akaike Information Criterion Correction (AICc) and statistically significant parameters were chosen as the best models in this research. Accordingly, for the new confirmed cases, ARIMA (3,1,14), ARIMA (0,1,11), ARIMA (1,0,10), and ARIMA (1,1,14) models were chosen for the UK, South Africa, Brazil, and Nigeria, respectively. Also, the model specification for the confirmed death cases was ARIMA (3,0,4), ARIMA (0,1,4), ARIMA (1,0,7), and ARIMA (Brown); models were selected for the UK, South Africa, Brazil, and Nigeria, respectively. The results of the ARIMA model forecasting showed that if the required measures are not taken by the respective governments and health practitioners in the days to come, the magnitude of the coronavirus pandemic is expected to increase in the study's selected countries.

Predicting regional COVID-19 hospital admissions in Sweden using mobility data.

The transmission of COVID-19 is dependent on social mixing, the basic rate of which varies with sociodemographic, cultural, and geographic factors. Alterations in social mixing and subsequent changes in transmission dynamics eventually affect hospital admissions. We employ these observations to model and predict regional hospital admissions in Sweden during the COVID-19 pandemic. We use an SEIR-model for each region in Sweden in which the social mixing is assumed to depend on mobility data from public transport utilisation and locations for mobile phone usage. The results show that the model could capture the timing of the first and beginning of the second wave of the pandemic 3 weeks in advance without any additional assumptions about seasonality. Further, we show that for two major regions of Sweden, models with public transport data outperform models using mobile phone usage. We conclude that a model based on routinely collected mobility data makes it possible to predict future hospital admissions for COVID-19 3 weeks in advance.

Predicting increases in COVID-19 incidence to identify locations for targeted testing in West Virginia: A machine learning enhanced approach.

During the COVID-19 pandemic, West Virginia developed an aggressive SARS-CoV-2 testing strategy which included utilizing pop-up mobile testing in locations anticipated to have near-term increases in SARS-CoV-2 infections. This study describes and compares two methods for predicting near-term SARS-CoV-2 incidence in West Virginia counties. The first method, Rt Only, is solely based on producing forecasts for each county using the daily instantaneous reproductive numbers, Rt. The second method, ML+Rt, is a machine learning approach that uses a Long Short-Term Memory network to predict the near-term number of cases for each county using epidemiological statistics such as Rt, county population information, and time series trends including information on major holidays, as well as leveraging statewide COVID-19 trends across counties and county population size. Both approaches used daily county-level SARS-CoV-2 incidence data provided by the West Virginia Department Health and Human Resources beginning April 2020. The methods are compared on the accuracy of near-term SARS-CoV-2 increases predictions by county over 17 weeks from January 1, 2021- April 30, 2021. Both methods performed well (correlation between forecasted number of cases and the actual number of cases week over week is 0.872 for the ML+Rt method and 0.867 for the Rt Only method) but differ in performance at various time points. Over the 17-week assessment period, the ML+Rt method outperforms the Rt Only method in identifying larger spikes. Results show that both methods perform adequately in both rural and non-rural predictions. Finally, a detailed discussion on practical issues regarding implementing forecasting models for public health action based on Rt is provided, and the potential for further development of machine learning methods that are enhanced by Rt.

Perceived e-learning stress as an independent predictor of e-learning readiness: Results from a nationwide survey in Bangladesh.

BACKGROUND: E-learning is a relatively trending system of education that has been placed over conventional campus-based learning worldwide, especially since the emergence of the COVID-19 pandemic. This study aimed to assess e-learning readiness among university students of a developing country like Bangladesh and identify the independent predictors of e-learning readiness. METHODS: From 26 December 2020 to 11 January 2021, a total of 1162 university students who had enrolled for e-learning completed a semi-structured questionnaire. Data were collected online via "Google Form" following the principles of snowball sampling through available social media platforms in Bangladesh. A multivariable linear regression model was fitted to investigate the association of e-learning readiness with perceived e-learning stress and other independent predictor variables. RESULTS: A total of 1162 university students participated in this study. The results indicated that with the increase of students' perceived e-learning stress score, the average e-learning readiness score was significantly decreased (ß = -0.43, 95% CI: -0.66, -0.20). The students did not seem ready, and none of the e-learning readiness scale items reached the highest mean score (5.0). The age, gender, divisional residence, preference of students and their parents, devices used, and having any eye problems were significantly associated with the students' e-learning readiness. CONCLUSION: During the prolonged period of the COVID-19 pandemic, e-learning implication strategies are needed to be assessed systematically with the level of readiness and its' impacts among students for the continuation of sound e-learning systems. The study findings recommend evaluating the e-learning readiness of university students and the mental health outcomes during the COVID-19 catastrophe in Bangladesh.

Estimating the COVID-19 prevalence and mortality using a novel data-driven hybrid model based on ensemble empirical mode decomposition.

In this study, we proposed a new data-driven hybrid technique by integrating an ensemble empirical mode decomposition (EEMD), an autoregressive integrated moving average (ARIMA), with a nonlinear autoregressive artificial neural network (NARANN), called the EEMD-ARIMA-NARANN model, to perform time series modeling and forecasting based on the COVID-19 prevalence and mortality data from 28 February 2020 to 27 June 2020 in South Africa and Nigeria. By comparing the accuracy level of forecasting measurements with the basic ARIMA and NARANN models, it was shown that this novel data-driven hybrid model did a better job of capturing the dynamic changing trends of the target data than the others used in this work. Our proposed mixture technique can be deemed as a helpful policy-supportive tool to plan and provide medical supplies effectively. The overall confirmed cases and deaths were estimated to reach around 176,570 [95% uncertainty level (UL) 173,607 to 178,476] and 3454 (95% UL 3384 to 3487), respectively, in South Africa, along with 32,136 (95% UL 31,568 to 32,641) and 788 (95% UL 775 to 804) in Nigeria on 12 July 2020 using this data-driven EEMD-ARIMA-NARANN hybrid technique. The contributions of this study include three aspects. First, the proposed hybrid model can better capture the dynamic dependency characteristics compared with the individual models. Second, this new data-driven hybrid model is constructed in a more reasonable way relative to the traditional mixture model. Third, this proposed model may be generalized to estimate the epidemic patterns of COVID-19 in other regions.

Time dynamics of COVID-19.

We apply tools from functional data analysis to model cumulative trajectories of COVID-19 cases across countries, establishing a framework for quantifying and comparing cases and deaths across countries longitudinally. It emerges that a country's trajectory during an initial first month "priming period" largely determines how the situation unfolds subsequently. We also propose a method for forecasting case counts, which takes advantage of the common, latent information in the entire sample of curves, instead of just the history of a single country. Our framework facilitates to quantify the effects of demographic covariates and social mobility on doubling rates and case fatality rates through a time-varying regression model. Decreased workplace mobility is associated with lower doubling rates with a roughly 2 week delay, and case fatality rates exhibit a positive feedback pattern.

Recommended reporting items for epidemic forecasting and prediction research: The EPIFORGE 2020 guidelines.

BACKGROUND: The importance of infectious disease epidemic forecasting and prediction research is underscored by decades of communicable disease outbreaks, including COVID-19. Unlike other fields of medical research, such as clinical trials and systematic reviews, no reporting guidelines exist for reporting epidemic forecasting and prediction research despite their utility. We therefore developed the EPIFORGE checklist, a guideline for standardized reporting of epidemic forecasting research. METHODS AND FINDINGS: We developed this checklist using a best-practice process for development of reporting guidelines, involving a Delphi process and broad consultation with an international panel of infectious disease modelers and model end users. The objectives of these guidelines are to improve the consistency, reproducibility, comparability, and quality of epidemic forecasting reporting. The guidelines are not designed to advise scientists on how to perform epidemic forecasting and prediction research, but rather to serve as a standard for reporting critical methodological details of such studies. CONCLUSIONS: These guidelines have been submitted to the EQUATOR network, in addition to hosting by other dedicated webpages to facilitate feedback and journal endorsement.

Possible future waves of SARS-CoV-2 infection generated by variants of concern with a range of characteristics.

Viral reproduction of SARS-CoV-2 provides opportunities for the acquisition of advantageous mutations, altering viral transmissibility, disease severity, and/or allowing escape from natural or vaccine-derived immunity. We use three mathematical models: a parsimonious deterministic model with homogeneous mixing; an age-structured model; and a stochastic importation model to investigate the effect of potential variants of concern (VOCs). Calibrating to the situation in England in May 2021, we find epidemiological trajectories for putative VOCs are wide-ranging and dependent on their transmissibility, immune escape capability, and the introduction timing of a postulated VOC-targeted vaccine. We demonstrate that a VOC with a substantial transmission advantage over resident variants, or with immune escape properties, can generate a wave of infections and hospitalisations comparable to the winter 2020-2021 wave. Moreover, a variant that is less transmissible, but shows partial immune-escape could provoke a wave of infection that would not be revealed until control measures are further relaxed.

Identifying and prioritizing potential human-infecting viruses from their genome sequences.

Determining which animal viruses may be capable of infecting humans is currently intractable at the time of their discovery, precluding prioritization of high-risk viruses for early investigation and outbreak preparedness. Given the increasing use of genomics in virus discovery and the otherwise sparse knowledge of the biology of newly discovered viruses, we developed machine learning models that identify candidate zoonoses solely using signatures of host range encoded in viral genomes. Within a dataset of 861 viral species with known zoonotic status, our approach outperformed models based on the phylogenetic relatedness of viruses to known human-infecting viruses (area under the receiver operating characteristic curve [AUC] = 0.773), distinguishing high-risk viruses within families that contain a minority of human-infecting species and identifying putatively undetected or so far unrealized zoonoses. Analyses of the underpinnings of model predictions suggested the existence of generalizable features of viral genomes that are independent of virus taxonomic relationships and that may preadapt viruses to infect humans. Our model reduced a second set of 645 animal-associated viruses that were excluded from training to 272 high and 41 very high-risk candidate zoonoses and showed significantly elevated predicted zoonotic risk in viruses from nonhuman primates, but not other mammalian or avian host groups. A second application showed that our models could have identified Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) as a relatively high-risk coronavirus strain and that this prediction required no prior knowledge of zoonotic Severe Acute Respiratory Syndrome (SARS)-related coronaviruses. Genome-based zoonotic risk assessment provides a rapid, low-cost approach to enable evidence-driven virus surveillance and increases the feasibility of downstream biological and ecological characterization of viruses.

Using machine learning for predicting intensive care unit resource use during the COVID-19 pandemic in Denmark.

The COVID-19 pandemic has put massive strains on hospitals, and tools to guide hospital planners in resource allocation during the ebbs and flows of the pandemic are urgently needed. We investigate whether machine learning (ML) can be used for predictions of intensive care requirements a fixed number of days into the future. Retrospective design where health Records from 42,526 SARS-CoV-2 positive patients in Denmark was extracted. Random Forest (RF) models were trained to predict risk of ICU admission and use of mechanical ventilation after n days (n = 1, 2, …, 15). An extended analysis was provided for n = 5 and n = 10. Models predicted n-day risk of ICU admission with an area under the receiver operator characteristic curve (ROC-AUC) between 0.981 and 0.995, and n-day risk of use of ventilation with an ROC-AUC between 0.982 and 0.997. The corresponding n-day forecasting models predicted the needed ICU capacity with a coefficient of determination (R2) between 0.334 and 0.989 and use of ventilation with an R2 between 0.446 and 0.973. The forecasting models performed worst, when forecasting many days into the future (for large n). For n = 5, ICU capacity was predicted with ROC-AUC 0.990 and R2 0.928, and use of ventilator was predicted with ROC-AUC 0.994 and R2 0.854. Random Forest-based modelling can be used for accurate n-day forecasting predictions of ICU resource requirements, when n is not too large.

A simple method to describe the COVID-19 trajectory and dynamics in any country based on Johnson cumulative density function fitting.

A simple method is utilised to study and compare COVID-19 infection dynamics between countries based on curve fitting to publicly shared data of confirmed COVID-19 infections. The method was tested using data from 80 countries from 6 continents. We found that Johnson cumulative density functions (CDFs) were extremely well fitted to the data (R2 > 0.99) and that Johnson CDFs were much better fitted to the tails of the data than either the commonly used normal or lognormal CDFs. Fitted Johnson CDFs can be used to obtain basic parameters of the infection wave, such as the percentage of the population infected during an infection wave, the days of the start, peak and end of the infection wave, and the duration of the wave's increase and decrease. These parameters can be easily interpreted biologically and used both for describing infection wave dynamics and in further statistical analysis. The usefulness of the parameters obtained was analysed with respect to the relation between the gross domestic product (GDP) per capita, the population density, the percentage of the population infected during an infection wave, the starting day and the duration of the infection wave in the 80 countries. We found that all the above parameters were significantly associated with GDP per capita, but only the percentage of the population infected was significantly associated with population density. If used with caution, this method has a limited ability to predict the future trajectory and parameters of an ongoing infection wave.