Complementing the power of deep learning with statistical model fusion: Probabilistic forecasting of influenza in Dallas County, Texas, USA

A Predictive Model of the Start of Annual Influenza Epidemics

Influenza is a respiratory disease that causes annual epidemics during cold seasons. These epidemics increase pressure on healthcare systems, sometimes provoking their collapse. For this reason, a tool is needed to predict when an influenza epidemic will occur so that the healthcare system has time to prepare for it. This study therefore aims to develop a statistical model capable of predicting the onset of influenza epidemics in Catalonia, Spain. Influenza seasons from 2011 to 2017 were used for model training, and those from 2017 to 2018 were used for validation. Logistic regression, Support Vector Machine, and Random Forest models were used to predict the onset of the influenza epidemic. The logistic regression model was able to predict the start of influenza epidemics at least one week in advance, based on clinical diagnosis rates of various respiratory diseases and meteorological variables. This model achieved the best punctual estimates for two of three performance metrics. The most important variables in the model were the principal components of bronchiolitis rates and mean temperature. The onset of influenza epidemics can be predicted from clinical diagnosis rates of various respiratory diseases and meteorological variables. Future research should determine whether predictive models play a key role in preventing influenza.

Multi-step influenza forecasting through singular value decomposition and kernel ridge regression with MARCOS-guided gradient-based optimization

This research delves into the significance of influenza outbreaks in public health, particularly the importance of accurate forecasts using weekly Influenza-like illness (ILI) rates. The present work develops a novel hybrid machine-learning model by combining singular value decomposition with kernel ridge regression (SKRR). In this context, a novel hybrid model known as H-SKRR is developed by combining two robust forecasting approaches, SKRR and ridge regression, which aims to improve multi-step-ahead predictions for weekly ILI rates in Southern and Northern China. The study begins with feature selection via XGBoost in the preprocessing phase, identifying optimal precursor information guided by importance factors. It decomposes the original signal using multivariate variational mode decomposition (MVMD) to address non-stationarity and complexity. H-SKRR is implemented by incorporating significant lagged-time components across sub-components. The aggregated forecasted values from these sub-components generate ILI values for two horizons (i.e., 4-and 7-weekly ahead). Employing the gradient-based optimization (GBO) algorithm fine-tunes model parameters. Furthermore, the deep random vector functional link (dRVFL), Ridge regression, and gated recurrent unit neural network (GRU) models were employed to validate the MVMD-H-SKRR-GBO paradigm's effectiveness. The outcomes, assessed using the MARCOS (Measurement of alternatives and ranking according to compromise solution) method as a multi-criteria decision-making method, highlight the superior accuracy of the MVMD-H-SKRR-GBO model in predicting ILI rates. The results clearly highlight the exceptional performance of the MVMD-H-SKRR-GBO model, with outstanding precision demonstrated by impressive R, RMSE, IA, and U95 % values of 0.946, 0.388, 0.970, and 1.075, respectively, at t + 7.

Integrating machine learning with otolith isoscapes: Reconstructing connectivity of a marine fish over four decades

Stable isotopes are an important tool to uncover animal migration. Geographic natal assignments often require categorizing the spatial domain through a nominal approach, which can introduce bias given the continuous nature of these tracers. Stable isotopes predicted over a spatial gradient (i.e., isoscapes) allow a probabilistic and continuous assignment of origin across space, although applications to marine organisms remain limited. We present a new framework that integrates nominal and continuous assignment approaches by (1) developing a machine-learning multi-model ensemble classifier using Bayesian model averaging (nominal); and (2) integrating nominal predictions with continuous isoscapes to estimate the probability of origin across the spatial domain (continuous). We applied this integrated framework to predict the geographic origin of the Northwest Atlantic mackerel (Scomber scombrus), a migratory pelagic fish comprised of northern and southern components that have distinct spawning sites off Canada (northern contingent) and the US (southern contingent), and seasonally overlap in the US fished regions. The nominal approach based on otolith carbon and oxygen stable isotopes (δ13C/δ18O) yielded high contingent classification accuracy (84.9%). Contingent assignment of unknown-origin samples revealed prevalent, yet highly varied contingent mixing levels (12.5-83.7%) within the US waters over four decades (1975-2019). Nominal predictions were integrated into mackerel-specific otolith oxygen isoscapes developed independently for Canadian and US waters. The combined approach identified geographic nursery hotspots in known spawning sites, but also detected geographic shifts over multi-decadal time scales. This framework can be applied to other marine species to understand migration and connectivity at a high spatial resolution, relevant to management of unit stocks in fisheries and other conservation assessments.

Application of big data and artificial intelligence in epidemic surveillance and containment

Faced with the current time-sensitive COVID-19 pandemic, the overburdened healthcare systems have resulted in a strong demand to develop newer methods to control the spread of the pandemic. Big data and artificial intelligence (AI) have been leveraged amid the COVID-19 pandemic; however, little is known about their use for supporting public health efforts. In epidemic surveillance and containment, efforts are needed to treat critical patients, track and manage the health status of residents, isolate suspected cases, and develop vaccines and antiviral drugs. The applications of emerging practices of artificial intelligence and big data have become powerful "weapons" to fight against the pandemic and provide strong support in pandemic prevention and control, such as early warning, analysis and judgment, interruption and intervention of epidemic, to achieve goals of early detection, early report, early diagnosis, early isolation and early treatment. These are the decisive factors to control the spread of the epidemic and reduce the mortality. This paper systematically summarized the application of big data and AI in epidemic, and describes practical cases and challenges with emphasis on epidemic prevention and control. The included studies showed that big data and AI have the potential strength to fight against COVID-19. However, many of the proposed methods are not yet widely accepted. Thus, the most rewarding research would be on methods that promise value beyond COVID-19. More efforts are needed for developing standardized reporting protocols or guidelines for practice.

Mathematical Modeling to Study Optimal Allocation of Vaccines against COVID-19 Using an Age-Structured Population

Vaccination against the coronavirus disease 2019 (COVID-19) started in early December of 2020 in the USA. The efficacy of the vaccines vary depending on the SARS-CoV-2 variant. Some countries have been able to deploy strong vaccination programs, and large proportions of their populations have been fully vaccinated. In other countries, low proportions of their populations have been vaccinated, due to different factors. For instance, countries such as Afghanistan, Cameroon, Ghana, Haiti and Syria have less than 10% of their populations fully vaccinated at this time. Implementing an optimal vaccination program is a very complex process due to a variety of variables that affect the programs. Besides, science, policy and ethics are all involved in the determination of the main objectives of the vaccination program. We present two nonlinear mathematical models that allow us to gain insight into the optimal vaccination strategy under different situations, taking into account the case fatality rate and age-structure of the population. We study scenarios with different availabilities and efficacies of the vaccines. The results of this study show that for most scenarios, the optimal allocation of vaccines is to first give the doses to people in the 55+ age group. However, in some situations the optimal strategy is to first allocate vaccines to the 15–54 age group. This situation occurs whenever the SARS-CoV-2 transmission rate is relatively high and the people in the 55+ age group have a transmission rate 50% or less that of those in the 15–54 age group. This study and similar ones can provide scientific recommendations for countries where the proportion of vaccinated individuals is relatively small or for future pandemics.

COVID 19 Pandemic, Socio-Economic Behaviour and Infection Characteristics: An Inter-Country Predictive Study Using Deep Learning

This work aims to develop a data driven multi-horizon incidence forecasting model considering the inter-country variability in static socio-economic factors. The specific objectives of this study are to predict the future country-wise COVID 19 incidences, to locate the influences of individual socio-economic factors on the predictions, to analyze the clusters of countries on the basis of influential explanatory variables and thus to search for intra-cluster and inter-cluster characteristics. To that respect this study has used the deep neural network based temporal fusion transformer for the predictions, Pearson correlation to understand the influence of socio-economic variables on incidence and hierarchical clustering for cluster-analysis. The findings conclude that the inter-country infection related predictions vary widely over spatio-temporal variability and different socio-economic variables have different influences over this inter-country variability. It is observed that greater the population size, stronger the global connectedness, larger the social cohesion, higher the population density and meaningful the gender based discrimination higher will be the future spread. On the other hand greater the development level, higher the nutritional status, greater the access to quality health services, greater the urban population and greater the material poverty lesser will be the future spread. Definite spatial pattern of influence of the explanatory variables emerged from cluster analysis. To minimize the vulnerability towards unforeseen biological calamities modern and sustainable development policies are needed; affluence may not guarantee less infection. But these policies should vary between economies due to the variation in socio-economic status of the countries worldwide.

Adaptively temporal graph convolution model for epidemic prediction of multiple age groups

Introduction

Multivariate time series prediction of infectious diseases is significant to public health, and the deep learning method has attracted increasing attention in this research field.

Material and methods

An adaptively temporal graph convolution (ATGCN) model, which learns the contact patterns of multiple age groups in a graph-based approach, was proposed for COVID- 19 and influenza prediction. We compared ATGCN with autoregressive models, deep sequence learning models, and experience- based ATGCN models in short-term and long-term prediction tasks.

Results

Results showed that the ATGCN model performed better than the autoregressive models and the deep sequence learning models on two datasets in both short-term (12.5% and 10% improvements on RMSE) and long-term (12.4% and 5% improvements on RMSE) prediction tasks. And the RMSE of ATGCN predictions fluctuated least in different age groups of COVID- 19 (0.029 ± 0.003) and influenza (0.059±0.008). Compared with the Ones-ATGCN model or the Pre-ATGCN model, the ATGCN model was more robust in performance, with RMSE of 0.0293 and 0.06 on two datasets when horizon is one.

Discussion

Our research indicates a broad application prospect of deep learning in the field of infectious disease prediction. Transmission characteristics and domain knowledge of infectious diseases should be further applied to the design of deep learning models and feature selection.

Conclusion

The ATGCN model addressed the multivariate time series forecasting in a graph-based deep learning approach and achieved robust prediction on the confirmed cases of multiple age groups, indicating its great potentials for exploring the implicit interactions of multivariate variables.

Analysis of Delayed Vaccination Regimens: A Mathematical Modeling Approach

The first round of vaccination against coronavirus disease 2019 (COVID-19) began in early December of 2020 in a few countries. There are several vaccines, and each has a different efficacy and mechanism of action. Several countries, for example, the United Kingdom and the USA, have been able to develop consistent vaccination programs where a great percentage of the population has been vaccinated (May 2021). However, in other countries, a low percentage of the population has been vaccinated due to constraints related to vaccine supply and distribution capacity. Countries such as the USA and the UK have implemented different vaccination strategies, and some scholars have been debating the optimal strategy for vaccine campaigns. This problem is complex due to the great number of variables that affect the relevant outcomes. In this article, we study the impact of different vaccination regimens on main health outcomes such as deaths, hospitalizations, and the number of infected. We develop a mathematical model of COVID-19 transmission to focus on this important health policy issue. Thus, we are able to identify the optimal strategy regarding vaccination campaigns. We find that for vaccines with high efficacy (>70%) after the first dose, the optimal strategy is to delay inoculation with the second dose. On the other hand, for a low first dose vaccine efficacy, it is better to use the standard vaccination regimen of 4 weeks between doses. Thus, under the delayed second dose option, a campaign focus on generating a certain immunity in as great a number of people as fast as possible is preferable to having an almost perfect immunity in fewer people first. Therefore, based on these results, we suggest that the UK implemented a better vaccination campaign than that in the USA with regard to time between doses. The results presented here provide scientific guidelines for other countries where vaccination campaigns are just starting, or the percentage of vaccinated people is small.

The roles of machine learning methods in limiting the spread of deadly diseases: A systematic review

Machine learning (ML) methods can be leveraged to prevent the spread of deadly infectious disease outbreak (e.g., COVID-19). This can be done by applying machine learning methods in predicting and detecting the deadly infectious disease. Most reviews did not discuss about the machine learning algorithms, datasets and performance measurements used for various applications in predicting and detecting the deadly infectious disease. In contrast, this paper outlines the literature review based on two major ways (e.g., prediction, detection) to limit the spread of deadly disease outbreaks. Hence, this study aims to investigate the state of the art, challenges and future works of leveraging ML methods to detect and predict deadly disease outbreaks according to two categories mentioned earlier. Specifically, this study provides a review on various approaches (e.g., individual and ensemble models), types of datasets, parameters or variables and performance measures used in the previous works. The literature review included all articles from journals and conference proceedings published from 2010 through 2020 in Scopus indexed databases using the search terms Predicting Disease Outbreaks and/or Detecting Disease using Machine Learning. The findings from this review focus on commonly used machine learning approaches, challenges and future works to limit the spread of deadly disease outbreaks through preventions and detections.

Nowcasting (Short-Term Forecasting) of Influenza Epidemics in Local Settings, Sweden, 2008-2019

The timing of influenza case incidence during epidemics can differ between regions within nations and states. We conducted a prospective 10-year evaluation (January 2008–February 2019) of a local influenza nowcasting (short-term forecasting) method in 3 urban counties in Sweden with independent public health administrations by using routine health information system data. Detection-of-epidemic-start (detection), peak timing, and peak intensity were nowcasted. Detection displayed satisfactory performance in 2 of the 3 counties for all nonpandemic influenza seasons and in 6 of 9 seasons for the third county. Peak-timing prediction showed satisfactory performance from the influenza season 2011–12 onward. Peak-intensity prediction also was satisfactory for influenza seasons in 2 of the counties but poor in 1 county. Local influenza nowcasting was satisfactory for seasonal influenza in 2 of 3 counties. The less satisfactory performance in 1 of the study counties might be attributable to population mixing with a neighboring metropolitan area.

Multi-step ahead meningitis case forecasting based on decomposition and multi-objective optimization methods

Epidemiological time series forecasting plays an important role in health public systems, due to its ability to allow managers to develop strategic planning to avoid possible epidemics. In this paper, a hybrid learning framework is developed to forecast multi-step-ahead (one, two, and three-month-ahead) meningitis cases in four states of Brazil. First, the proposed approach applies an ensemble empirical mode decomposition (EEMD) to decompose the data into intrinsic mode functions and residual components. Then, each component is used as the input of five different forecasting models, and, from there, forecasted results are obtained. Finally, all combinations of models and components are developed, and for each case, the forecasted results are weighted integrated (WI) to formulate a heterogeneous ensemble forecaster for the monthly meningitis cases. In the final stage, a multi-objective optimization (MOO) using the Non-Dominated Sorting Genetic Algorithm – version II is employed to find a set of candidates’ weights, and then the Technique for Order of Preference by similarity to Ideal Solution (TOPSIS) is applied to choose the adequate set of weights. Next, the most adequate model is the one with the best generalization capacity out-of-sample in terms of performance criteria including mean absolute error (MAE), relative root mean squared error (RRMSE), and symmetric mean absolute percentage error (sMAPE). By using MOO, the intention is to enhance the performance of the forecasting models by improving simultaneously their accuracy and stability measures. To access the model’s performance, comparisons based on metrics are conducted with: (i) EEMD, heterogeneous ensemble integrated by direct strategy, or simple sum; (ii) EEMD, homogeneous ensemble of components WI; (iii) models without signal decomposition. At this stage, MAE, RRMSE, and sMAPE criteria as well as Diebold–Mariano statistical test are adopted. In all twelve scenarios, the proposed framework was able to perform more accurate and stable forecasts, which showed, on 89.17% of the cases, that the errors of the proposed approach are statistically lower than other approaches. These results showed that combining EEMD, heterogeneous ensemble and WI with weights obtained by optimization can develop precise and stable forecasts. The modeling developed in this paper is promising and can be used by managers to support decision making.

•

Ensemble empirical mode decomposition is applied into the raw time series.

•

Heterogeneous ensemble learning models are used to forecasting meningitis cases.

•

The NSGA-II algorithm and TOPSIS criterion are employed in the multi-objective procedure.

•

Proposed model has errors statistically lower than 89.17% of the compared models.

•

Promising results are achieved by the weighted integrated ensemble learning model.

Using deep learning to predict the hand-foot-and-mouth disease of enterovirus A71 subtype in Beijing from 2011 to 2018

Hand-foot-and-month disease (HFMD), especially the enterovirus A71 (EV-A71) subtype, is a major health problem in Beijing, China. Previous studies mainly used regressive models to forecast the prevalence of HFMD, ignoring its intrinsic age groups. This study aims to predict HFMD of EV-A71 subtype in three age groups (0–3, 3–6 and > 6 years old) from 2011 to 2018 using residual-convolutional-recurrent neural network (CNNRNN-Res), convolutional-recurrent neural network (CNNRNN) and recurrent neural network (RNN). They were compared with auto-regressio, global auto-regression and vector auto-regression on both short-term and long-term prediction. Results showed that CNNRNN-Res and RNN had higher accuracies on point forecast tasks, as well as robust performances in long-term prediction. Three deep learning models also had better skills in peak intensity forecast, and CNNRNN-Res achieved the best results in the peak month forecast. We also found that three age groups had consistent outbreak trends and similar patterns of prediction errors. These results highlight the superior performance of deep learning models in HFMD prediction and can assist the decision-makers to refine the HFMD control measures according to age groups.