key: cord-0926628-3vm8kfc4 authors: Ali, D. A.; Midi, H. title: Short-term Forecasting of Cumulative Confirmed Cases of Covid-19 Pandemic in Somalia date: 2020-06-20 journal: nan DOI: 10.1101/2020.06.18.20135053 sha: 1cdca548346546aed18539b7c828868d0c61cc74 doc_id: 926628 cord_uid: 3vm8kfc4 Somalia has recorded the first confirmed Covid-19 case and first death case on March 16, and April 08, 2020, respectively. Since its arrival to the country, it had infected 2,603 people and took the lives of 88 people while 577 patients were recovered as of 14 June, 2020. To fight this pandemic, the government requires to make the necessary plans accordingly. To plan effectively, the government needs to answer this question: what will be the effect of Covid-19 cases in the country? To answer this question accurately and objectively, forecasting the spread of confirmed Covid-19 cases will be vital.To this regard, this paper provides real times forecasts of Covid-19 cases employing Holt's linear trend model without seasonality. Provided that the data employed is accurate and the past pattern of the disease will continue in the future, this model is powerful to produce real time forecasts in the future with some degree of uncertainty. With the help of these forecasts,the government can make evidence based decisions by utilizing the scarce resource available at its disposal. Covid-19 pandemic is a communicable disease which resulted from the severe acute respiratory syndrome. Although it first started in December 2019 in Wuhan city, in China, it is currently world-wide epidemic as declared by World Health Organization 1 . It currently infected more than 7.5 million people and killed more than 428,000 people globally 2 as of June 13,2020 . Although the governments across the globe implemented some containment measures such as curfews, wearing masks, hand sanitization and so on to fight against this pandemic, the number of confirmed cases continued to show an increasing trend. While it imposed a serious threat to the public health sector, other sectors such as economies were also significantly affected 3, 4 . Fever, shortness of breath as well as dry cough are the common symptoms of this Covid-19 pandemic 5 while mild diarrhea, loss of smell, sore throat and muscle pain are among the uncommon signs of the novel coronavirus. The Covid-19 pandemic is transmitted through respiratory droplets which are generated when an individual sneezes or coughs. Also, a person can get infected by touching a contaminated surface or coming into a contact with an infected person 6 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 20, 2020. Simple time series forecasting techniques were employed to forecast the cumulative confirmed cases of Covid-19 pandemic. Models from exponential smoothing family were adopted to produce reliable forecasts 9, 10 . According to 11, 12 , these models provide forecasts with good precisions relative to its competitors especially when the sample size is small. These According to Fig1, the cumulative confirmed cases experienced an exponential pattern with increasing trend. The curve has not yet started to flatten or has not yet reached a plateau while a gradual surge was observed in number of recovered cases. First strand of forecasts: From May 11 to May 20, 2020. We began our June 19, 2020 4/11 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 20, 2020. . first set of forecasts at the end May 10, 2020 with thirty-three observations. The forecasted cumulative cases with 95% confidence intervals were produced at the end of May 10, 2020 are depicted in Table 1 . The forecasted mean for cumulative confirmed cases 10-days-step was 1,521 with 95% prediction interval which ranged between 1,234 and 1,808 cases while the observed cumulative cases on May 20, 2020 was 1,573 cases. An absolute percent error of 3.3% was observed on May 20, 2020. Nevertheless, the observed cumulative cases remained within the predication interval throughout this period. A mean absolute percent error of 0.73%, a measure of forecasting accuracy, was observed. Second strand of forecasts: From May 21 to May 30, 2020. We updated our data set to include cumulative confirmed cases up to May 20, 2020 and again 10 days-step forecasts were generated. The mean estimates with 95% prediction intervals are shown in Table 2 . The forecasted cumulative cases June 19, 2020 5/11 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 20, 2020. . An absolute percent error of 9.5%(with the point forecast being slightly positively biased) was observed which is higher than the previous 10 days of forecasts. However, the forecasts still fall within the prediction interval. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20135053 doi: medRxiv preprint forecasted cumulative confirmed cases on June 09, 2020 were 2,331 cases while the actual confirmed cases were 2,409 cases. This time round, we recorded an absolute percent forecast error of 3.2 percent which was noticeably lower than previous forecasting error of 9.5 percent. This means that the forecasting uncertainty was higher in the preceding strand of forecasts. June 19, 2020 8/11 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20135053 doi: medRxiv preprint To forecast the cumulative confirmed cases in Somalia, we employed an exponential smoothing family especially holt winter's linear model without seasonality. This model like the rest of forecasting models assume that whatever measures or policies the government took to fight this Covid-19 pandemic in the past will continue to exist in the future. On top of that the accuracy of data used to produce these forecasts is also assumed. In this study, we produced four strands of forecasts and the first three covered the period from May 11 to June 09, 2020. In the first and the third strands of forecasts, the observed and estimated cumulative confirmed cases were near to each other which recorded an average absolute percent error of 0.73 percent and 0.75 percent, respectively. On the contrary, the second set of forecasts recorded an average absolute percent error of 8.7% percent and this implies that the realized and forecasted cumulative confirmed cases were not close as in the first and the third strands of forecasts. The gist of this study was to help the government make necessary plans accordingly. This model provides real time forecasting which can help the government and other stakeholders to get a clear picture on where the maximum number of Covid-19 cases in the country could reach in the future with some uncertainty. With help of these forecasts, the government can make evidence based decisions by utilizing the scarce resource available at its disposal. To this end, forecasting should be the heart and an integral part in the decision making process specifically when the government is dealing with a pandemic like Covid-19. This Covid-19 epidemic did not only pose threat to the public health sector but also severely affected many other sectors such as the economy of the country. June 19, 2020 9/11 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20135053 doi: medRxiv preprint COVID-19 CORONAVIRUS PANDEMIC 2020 How coronavirus could impact the global supply chain by mid-march JPMorgan officially forecasts a coronavirus-driven recession will rock the US and Europe by A new coronavirus associated with human respiratory disease in China Early dynamics of transmission and control of COVID-19: a mathematical modelling study. The lancet infectious diseases Global surveillance for COVID-19 caused by human infection Ministry of Health Somalia. COVID-19 Situational Update A state space framework for automatic forecasting using exponential smoothing methods Exponential smoothing with a damped multiplicative trend The accuracy of extrapolation (time series) methods: Results of a forecasting competition The M4 Competition: 100,000 time series and 61 forecasting methods Exponential smoothing models: Means and variances for lead-time demand