Dataset Information

Time series prediction of under-five mortality rates for Nigeria: comparative analysis of artificial neural networks, Holt-Winters exponential smoothing and autoregressive integrated moving average models.

ABSTRACT:

Background

Accurate forecasting model for under-five mortality rate (U5MR) is essential for policy actions and planning. While studies have used traditional time series modeling techniques (e.g., autoregressive integrated moving average (ARIMA) and Holt-Winters smoothing exponential methods), their appropriateness to predict noisy and non-linear data (such as childhood mortality) has been debated. The objective of this study was to model long-term U5MR with group method of data handling (GMDH)-type artificial neural network (ANN), and compare the forecasts with the commonly used conventional statistical methods-ARIMA regression and Holt-Winters exponential smoothing models.

Methods

The historical dataset of annual U5MR in Nigeria from 1964 to 2017 was obtained from the official website of World Bank. The optimal models for each forecasting methods were used for forecasting mortality rates to 2030 (ending of Sustainable Development Goal era). The predictive performances of the three methods were evaluated, based on root mean squared errors (RMSE), root mean absolute error (RMAE) and modified Nash-Sutcliffe efficiency (NSE) coefficient. Statistically significant differences in loss function between forecasts of GMDH-type ANN model compared to each of the ARIMA and Holt-Winters models were assessed with Diebold-Mariano (DM) test and Deming regression.

Results

The modified NSE coefficient was slightly lower for Holt-Winters methods (96.7%), compared to GMDH-type ANN (99.8%) and ARIMA (99.6%). The RMSE of GMDH-type ANN (0.09) was lower than ARIMA (0.23) and Holt-Winters (2.87). Similarly, RMAE was lowest for GMDH-type ANN (0.25), compared with ARIMA (0.41) and Holt-Winters (1.20). From the DM test, the mean absolute error (MAE) was significantly lower for GMDH-type ANN, compared with ARIMA (difference?=?0.11, p-value?=?0.0003), and Holt-Winters model (difference?=?0.62, p-value0 =?0.004?±?standard error: 0.06; 95% confidence interval: -?0.113 to 0.122).

Conclusions

GMDH-type neural network performed better in predicting and forecasting of under-five mortality rates for Nigeria, compared to the ARIMA and Holt-Winters models. Therefore, GMDH-type ANN might be more suitable for data with non-linear or unknown distribution, such as childhood mortality. GMDH-type ANN increases forecasting accuracy of childhood mortalities in order to inform policy actions in Nigeria.

SUBMITTER: Adeyinka DA

PROVIDER: S-EPMC7712624 | biostudies-literature | 2020 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Time series prediction of under-five mortality rates for Nigeria: comparative analysis of artificial neural networks, Holt-Winters exponential smoothing and autoregressive integrated moving average models.

Adeyinka Daniel Adedayo DA Muhajarine Nazeem N

BMC medical research methodology 20201203 1

<h4>Background</h4>Accurate forecasting model for under-five mortality rate (U5MR) is essential for policy actions and planning. While studies have used traditional time series modeling techniques (e.g., autoregressive integrated moving average (ARIMA) and Holt-Winters smoothing exponential methods), their appropriateness to predict noisy and non-linear data (such as childhood mortality) has been debated. The objective of this study was to model long-term U5MR with group method of data handling ...[more]

PMID: 33267817

Similar Datasets

Project description:The coronavirus disease 2019 (COVID-19) spread rapidly across the world since its appearance in December 2019. This data set creates one-, three-, and seven-day forecasts of the COVID-19 pandemic's cumulative case counts at the county, health district, and state geographic levels for the state of Virginia. Forecasts are created over the first 46 days of reported COVID-19 cases using the cumulative case count data provided by The New York Times as of April 22, 2020. From this historical data, one-, three-, seven, and all-days prior to the forecast start date are used to generate the forecasts. Forecasts are created using: (1) a Naïve approach; (2) Holt-Winters exponential smoothing (HW); (3) growth rate (Growth); (4) moving average (MA); (5) autoregressive (AR); (6) autoregressive moving average (ARMA); and (7) autoregressive integrated moving average (ARIMA). Median Absolute Error (MdAE) and Median Absolute Percentage Error (MdAPE) metrics are created with each forecast to evaluate the forecast with respect to existing historical data. These error metrics are aggregated to provide a means for assessing which combination of forecast method, forecast length, and lookback length are best fits, based on lowest aggregated error at each geographic level. The data set is comprised of an R-Project file, four R source code files, all 1,329,404 generated short-range forecasts, MdAE and MdAPE error metric data for each forecast, copies of the input files, and the generated comparison tables. All code and data files are provided to provide transparency and facilitate replicability and reproducibility. This package opens directly in RStudio through the R Project file. The R Project file removes the need to set path locations for the folders contained within the data set to simplify setup requirements. This data set provides two avenues for reproducing results: 1) Use the provided code to generate the forecasts from scratch and then run the analyses; or 2) Load the saved forecast data and run the analyses on the stored data. Code annotations provide the instructions needed to accomplish both routes. This data can be used to generate the same set of forecasts and error metrics for any US state by altering the state parameter within the source code. Users can also generate health district forecasts for any other state, by providing a file which maps each county within a state to its respective health-district. The source code can be connected to the most up-to-date version of The New York Times COVID-19 dataset allows for the generation of forecasts up to the most recently reported data to facilitate near real-time forecasting.

Project description:BACKGROUND:Measuring hepatic R2* by fitting a monoexponential model to the signal decay of a multigradient-echo (mGRE) sequence noninvasively determines hepatic iron content (HIC). Concurrent hepatic steatosis introduces signal oscillations and confounds R2* quantification with standard monoexponential models. PURPOSE:To evaluate an autoregressive moving average (ARMA) model for accurate quantification of HIC in the presence of fat using biopsy as the reference. STUDY TYPE:Phantom study and in vivo cohort. POPULATION:Twenty iron-fat phantoms covering clinically relevant R2* (30-800?s-1 ) and fat fraction (FF) ranges (0-40%), and 10 patients (four male, six female, mean age 18.8?years). FIELD STRENGTH/SEQUENCE:2D mGRE acquisitions at 1.5?T and 3?T. ASSESSMENT:Phantoms were scanned at both field strengths. In vivo data were analyzed using the ARMA model to determine R2* and FF values, and compared with biopsy results. STATISTICAL TESTS:Linear regression analysis was used to compare ARMA R2* and FF results with those obtained using a conventional monoexponential model, complex-domain nonlinear least squares (NLSQ) fat-water model, and biopsy. RESULTS:In phantoms and in vivo, all models produced R2* and FF values consistent with expected values in low iron and low/high fat conditions. For high iron and no fat phantoms, monoexponential and ARMA models performed excellently (slopes: 0.89-1.07), but NLSQ overestimated R2* (slopes: 1.14-1.36) and produced false FFs (12-17%) at 1.5?T; in high iron and fat phantoms, NLSQ (slopes: 1.02-1.16) outperformed monoexponential and ARMA models (slopes: 1.23-1.88). The results with NLSQ and ARMA improved in phantoms at 3?T (slopes: 0.96-1.04). In patients, mean R2*-HIC estimates for monoexponential and ARMA models were close to biopsy-HIC values (slopes: 0.90-0.95), whereas NLSQ substantially overestimated HIC (slope 1.4) and produced false FF values (4-28%) with very high SDs (15-222%) in patients with high iron overload and no steatosis. DATA CONCLUSION:ARMA is superior in quantifying R2* and FF under high iron and no fat conditions, whereas NLSQ is superior for high iron and concurrent fat at 1.5?T. Both models give improved R2* and FF results at 3?T. LEVEL OF EVIDENCE:2 Technical Efficacy Stage: 2 J. Magn. Reson. Imaging 2019;50:1620-1632.

Project description:Influenza A virus commonly circulating in swine (IAV-S) is characterized by large genetic and antigenic diversity and, thus, improvements in different aspects of IAV-S surveillance are needed to achieve desirable goals of surveillance such as to establish the capacity to forecast with the greatest accuracy the number of influenza cases likely to arise. Advancements in modeling approaches provide the opportunity to use different models for surveillance. However, in order to make improvements in surveillance, it is necessary to assess the predictive ability of such models. This study compares the sensitivity and predictive accuracy of the autoregressive integrated moving average (ARIMA) model, the generalized linear autoregressive moving average (GLARMA) model, and the random forest (RF) model with respect to the frequency of influenza A virus (IAV) in Ontario swine. Diagnostic data on IAV submissions in Ontario swine between 2007 and 2015 were obtained from the Animal Health Laboratory (University of Guelph, Guelph, ON, Canada). Each modeling approach was examined for predictive accuracy, evaluated by the root mean square error, the normalized root mean square error, and the model's ability to anticipate increases and decreases in disease frequency. Likewise, we verified the magnitude of improvement offered by the ARIMA, GLARMA and RF models over a seasonal-naïve method. Using the diagnostic submissions, the occurrence of seasonality and the long-term trend in IAV infections were also investigated. The RF model had the smallest root mean square error in the prospective analysis and tended to predict increases in the number of diagnostic submissions and positive virological submissions at weekly and monthly intervals with a higher degree of sensitivity than the ARIMA and GLARMA models. The number of weekly positive virological submissions is significantly higher in the fall calendar season compared to the summer calendar season. Positive counts at weekly and monthly intervals demonstrated a significant increasing trend. Overall, this study shows that the RF model offers enhanced prediction ability over the ARIMA and GLARMA time series models for predicting the frequency of IAV infections in diagnostic submissions.

Project description:BackgroundR2*-MRI is clinically used to noninvasively assess hepatic iron content (HIC) to guide potential iron chelation therapy. However, coexisting pathologies, such as fibrosis and steatosis, affect R2* measurements and may thus confound HIC estimations.PurposeTo evaluate whether a multispectral auto regressive moving average (ARMA) model can be used in conjunction with quantitative susceptibility mapping (QSM) to measure magnetic susceptibility as a confounder-free predictor of HIC.Study typePhantom study and in vivo cohort.SubjectsNine iron phantoms covering clinically relevant R2* range (20-1200/second) and 48 patients (22 male, 26 female, median age 18 years).Field strength/sequenceThree-dimensional (3D) and two-dimensional (2D) multi-echo gradient echo (GRE) at 1.5 T.AssessmentARMA-QSM modeling was performed on the complex 3D GRE signal to estimate R2*, fat fraction (FF), and susceptibility measurements. R2*-based dry clinical HIC values were calculated from the 2D GRE acquisition using a published R2*-HIC calibration curve as reference standard.Statistical testsLinear regression analysis was performed to compare ARMA R2* and susceptibility-based estimates to iron concentrations and dry clinical HIC values in phantoms and patients, respectively.ResultsIn phantoms, the ARMA R2* and susceptibility values strongly correlated with iron concentrations (R2 ≥ 0.9). In patients, the ARMA R2* values highly correlated (R2 = 0.97) with clinical HIC values with slope = 0.026, and the susceptibility values showed good correlation (R2 = 0.82) with clinical dry HIC values with slope = 3.3 and produced a dry-to-wet HIC ratio of 4.8.Data conclusionThis study shows the feasibility that ARMA-QSM can simultaneously estimate susceptibility-based wet HIC, R2*-based dry HIC and FFs from a single multi-echo GRE acquisition. Our results demonstrate that both, R2* and susceptibility-based wet HIC values estimated with ARMA-QSM showed good association with clinical dry HIC values with slopes similar to published R2*-biopsy HIC calibration and dry-to-wet tissue weight ratio, respectively. Hence, our study shows that ARMA-QSM can provide potentially confounder-free assessment of hepatic iron overload.Level of evidence3 TECHNICAL EFFICACY: Stage 2.

Project description:BackgroundExisting studies analyzing the impact of state concussion laws have found an increase in concussion-related medical encounters post-law, in some instances, such increases were observed during the pre-law period due to a potential "spillover" effect. This study assessed the effects of Ohio's concussion law, while accounting for such a "spillover" effect, on the trends in monthly rates of concussion-related medical encounters in Medicaid insured children using autoregressive integrated moving average (ARIMA) analysis.MethodsWe analyzed claim data obtained from the Partners For Kids database, a pediatric accountable care organization in Ohio. Concussion-related medical encounters for Medicaid-insured children (ages 0-18 years) treated between April 1, 2008 to December 31, 2016 were selected and analyzed. We assessed pre- and post-law trends in concussion-related medical encounters using an ARIMA intervention model. We also used traditional regression methods to validate the study results.ResultsA total of 16,943 concussion-related medical encounters sustained by 15,545 unique patients were included. Monthly rates of concussion-related medical encounters significantly increased from 4.64 per 10,000 member months during the pre-law period to 6.69 per 10,000 member months in the post-law period (P < 0.0001). Three upward breaks in the monthly rates of concussion-related medical encounters were observed between 2009 and 2016, with two breaks observed during the pre-law period. Specifically, the increased breakpoint observed in July 2011 (P = 0.0186) was significantly associated with an estimated 7.3% increase (95% CI: 1.1-13.7) in the rate of concussion-related medical encounters. This finding was confirmed in the Poisson regression and curve fitting models. Furthermore, a seasonal trend in concussion-related medical encounters was observed with the highest rates in September and October of each year.ConclusionsTwo of the three upward breaks identified in the monthly rate of concussion-related medical encounters occurred before the enactment of Ohio's concussion law, suggesting a potential "spillover" effect. Further research is needed to confirm such an effect in children with other types of medical insurance.

Project description:BackgroundAlthough vaccination is one of the main countermeasures against influenza epidemic, it is highly essential to make informed prevention decisions to guarantee that limited vaccination resources are allocated to the places where they are most needed. Hence, one of the fundamental steps for decision making in influenza prevention is to characterize its spatio-temporal trend, especially on the key problem about how influenza transmits among adjacent places and how much impact the influenza of one place could have on its neighbors. To solve this problem while avoiding too much additional time-consuming work on data collection, this study proposed a new concept of spatio-temporal route as well as its estimation methods to construct the influenza transmission network.MethodsThe influenza-like illness (ILI) data of Sichuan province in 21 cities was collected from 2010 to 2016. A joint pattern based on the dynamic Bayesian network (DBN) model and the vector autoregressive moving average (VARMA) model was utilized to estimate the spatio-temporal routes, which were applied to the two stages of learning process respectively, namely structure learning and parameter learning. In structure learning, the first-order conditional dependencies approximation algorithm was used to generate the DBN, which could visualize the spatio-temporal routes of influenza among adjacent cities and infer which cities have impacts on others in influenza transmission. In parameter learning, the VARMA model was adopted to estimate the strength of these impacts. Finally, all the estimated spatio-temporal routes were put together to form the final influenza transmission network.ResultsThe results showed that the period of influenza transmission cycle was longer in Western Sichuan and Chengdu Plain than that in Northeastern Sichuan, and there would be potential spatio-temporal routes of influenza from bordering provinces or municipalities into Sichuan province. Furthermore, this study also pointed out several estimated spatio-temporal routes with relatively high strength of associations, which could serve as clues of hot spot areas detection for influenza surveillance.ConclusionsThis study proposed a new framework for exploring the potentially stable spatio-temporal routes between different places and measuring specific the sizes of transmission effects. It could help making timely and reliable prediction of the spatio-temporal trend of infectious diseases, and further determining the possible key areas of the next epidemic by considering their neighbors' incidence and the transmission relationships.

Project description:BackgroundMathematical and statistical models are used to predict trends in epidemic spread and determine the effectiveness of control measures. Automatic regressive integrated moving average (ARIMA) models are used for time-series forecasting, but only few models of the 2019 coronavirus disease (COVID-19) pandemic have incorporated protective behaviors or vaccination, known to be effective for pandemic control.MethodsTo improve the accuracy of prediction, we applied newly developed ARIMA models with predictors (mask wearing, avoiding going out, and vaccination) to forecast weekly COVID-19 case growth rates in Canada, France, Italy, and Israel between January 2021 and March 2022. The open-source data was sourced from the YouGov survey and Our World in Data. Prediction performance was evaluated using the root mean square error (RMSE) and the corrected Akaike information criterion (AICc).ResultsA model with mask wearing and vaccination variables performed best for the pandemic period in which the Alpha and Delta viral variants were predominant (before November 2021). A model using only past case growth rates as autoregressive predictors performed best for the Omicron period (after December 2021). The models suggested that protective behaviors and vaccination are associated with the reduction of COVID-19 case growth rates, with booster vaccine coverage playing a particularly vital role during the Omicron period. For example, each unit increase in mask wearing and avoiding going out significantly reduced the case growth rate during the Alpha/Delta period in Canada (-0.81 and -0.54, respectively; both p < 0.05). In the Omicron period, each unit increase in the number of booster doses resulted in a significant reduction of the case growth rate in Canada (-0.03), Israel (-0.12), Italy (-0.02), and France (-0.03); all p < 0.05.ConclusionsThe key findings of this study are incorporating behavior and vaccination as predictors led to accurate predictions and highlighted their significant role in controlling the pandemic. These models are easily interpretable and can be embedded in a "real-time" schedule with weekly data updates. They can support timely decision making about policies to control dynamically changing epidemics.

Dataset Information

Time series prediction of under-five mortality rates for Nigeria: comparative analysis of artificial neural networks, Holt-Winters exponential smoothing and autoregressive integrated moving average models.

Background

Methods

Results

Conclusions

Publications

Time series prediction of under-five mortality rates for Nigeria: comparative analysis of artificial neural networks, Holt-Winters exponential smoothing and autoregressive integrated moving average models.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets