Accurate prediction of cumulative COVID-19 infected cases is essential for effectively managing the limited healthcare resources in India. Historically, epidemiological models have helped in controlling such epidemics. Models require accurate historical data to predict future outcomes. In our data, there were days exhibiting erratic, apparently anomalous jumps and drops in the number of daily reported COVID-19 infected cases that did not conform with the overall trend. Including those observations in the training data would most likely worsen model predictive accuracy. However, with existing epidemiological models it is not straightforward to determine, for a specific day, whether or not an outcome should be considered anomalous. In this work, we propose an algorithm to automatically identify anomalous ‘jump’ and ‘drop’ days, and then based upon the overall trend, the number of daily infected cases for those days is adjusted and the training data is amended using the adjusted observations. We applied the algorithm in conjunction with a recently proposed, modified Susceptible-Infected-Susceptible (SIS) model to demonstrate that prediction accuracy is improved after adjusting training data counts for apparent erratic anomalous jumps and drops.

Accurate prediction of cumulative COVID-19 infected cases is essential for effectively managing the limited healthcare resources, specifically in India. Historically, epidemiological models have helped control such epidemics. Those models need accurate past data to predict for future. However, for various reasons, we observe sudden drops and jumps in the number of daily reported COVID-19 infected cases on some days, not aligned with the overall trend. If we incorporate those observations in the training data, the model's prediction accuracy may worsen, as they do not capture the correct trend in the training data. But it is not straightforward for the existing epidemiological models to decide a specific day as a sudden drop or jump. We propose an algorithm that automatically determines any drop or jump days in this work. Then, based on the overall trend in the data, we adjust the number of daily infected cases on those days and decide on the training data based on the adjusted observations. We have applied the proposed algorithm in a recently proposed modified Susceptible-Infected-Susceptible (SIS) to show that adjusted training data gives better prediction accuracy when jump and drop exist in the training data.