AI powered data analytics for Corona Virus epidemic monitoring and control.

As the first few cases emerged in Wuhan, around the 31st of December information had reached the WHO. Early stages showed pneumonia like syndrome and had heavy similarities with the SARs virus. Whenever such a situation arises, we have very less or no clue at all that this might turn into a global disaster. By the time the severity and destructive strength of the virus was detected it had already flown to various countries and had started the pandemic. As an early protective measure government across the world had stopped the air transport facilities to a certain extend and had advised citizens to stay quarantined. But the big picture here is that these implemented procedures are no fool proof solutions to tackle the situation.

Now to tackle such extensively spreading diseases or situations it would require heavy data rendering and processing artificial intelligence system working at various levels, that are able to predict and foresee the damage the disease can cause and design solutions to deal with it. If the AI system were to predict the intensity of the spread rate at a very early stage and proposed the right lockdown period, the pandemic control would have been a success. But while the design of such an artificial intelligence system that promises on predicting a global disaster, it would have to be very accurate as it is responsible for many lives. But building a 100% accurate artificial intelligence model is rarely possible, but AI models with slightly reduced accuracy rates have been announced in the recent times. Alibaba was supposed to release an AI model with about 96% accuracy which would categorise CT scans of pneumonia patients as infected and non-infected. Similarly, US military had announced an AI algorithm that could predict infection in non-symptomatic patients before the signs are visible with an accuracy of about 85%. Canadian company blue dot said to have observed signals of a potential pandemic long before the actual surge and had warned the officials well before hand. These signals where picked up by sophisticated AI models. AI has the capacity to work with enormous amounts of continuously updating data, and to render it and to produce meaningful conclusions.

The interesting thing to note here is the possibility of detecting a potential pandemic from the macro data collected from the society from invariably everywhere. And this would not produce a security threat as these macro data collected from the user serves the purpose of understanding the mood and flavour of the society. Micro and microdata would include social media posts, amount of people visiting the doctor, symptom details, cash flow details and its rate, amount and places of ATM swipes, flight and transport details. These data could be integrated and fed into an AI model which according to the given data could analyse the mood of a certain society and predict the possibility of an upcoming pandemic. Sudden changes in patterns of domestic and international flights, direction and intensity of traffic flow, increase in procurement of medical supplies, retail patterns and increasing sentiments of a certain disease on social media suggests a peculiar event occurring in the community. Now there will be multiple layered AI model that would be used to first predict the possibility of a potential disease then further layered AI would be used to predict the damages it could cause and further layered AI would be used to provide a suitable defence mechanism to the problem.

In multi-layered complex AI algorithm once the chance of a potential pandemic is observed it starts working on the fate of covid 19 patients. The model would be fed with medical data containing all the symptoms that are shown. Now the algorithm would decide the fate of the patient with the symptoms mentioned. The patient could show various symptoms but may not necessarily be in fatal condition, but a patient showing only a few symptoms could turn serious. Now the AI model did not find most symptoms leading to a confirmed fatality, but it was able to narrow down 3 symptoms that ensured serious infection with heavy accuracy. Increase in haemoglobin levels, muscle aching(myglia) and the subtle fluctuation of presence of enzyme called alanine aminotransferase in the liver ensured the patient to be seriously infected. The prediction model also involved data feeding from a swarm of people and later analysing using probabilities, there is random forest model where all patient details are fed and different optional trees are created and finally the most accurate tree is selected that could perform the most accurate prediction of the fate of the covid infected patient.

Now considering the seriousness of the role performed by the AI system it is important to consider that the most accurate model must be used. A separate committee was designated in order to find the most preferable and reliable artificial intelligence model. Now there where 5 different groups of approaches that an AI model could use to produce predictions. Now these groups were evaluated according to their level of complexity, time taken to produce a certain conclusion, accuracy of the prediction is monitored. Clinical surveillance and the predictive surveillance of the patient data is matched and scored for accuracy. There could be basically 3 sets of models, it could be multi input models, parametric models or non-parametric models. Now a set of predefined small data is passed within these algorithms and the accuracy % is matched, now finally in panel selection the winner is selected and invariably the winner produces the most accurate prediction. This method is also known as the GROOMS methodology of selection of suitable AI model from all 5 different groups with predefined sets of characteristics.

Now one of the most important prediction that the AI model will have to make is the spread rate and possible infection probability strength of the virus. These predictions are made using the SIR and SEIR AI model. These are heavily mathematically modelled data which would finally produce a destined result that would predict the possible rate of spread of the virus in a particular community. Now this prediction necessarily would have very little accuracy as this prediction solely relies on the historical spread rate data that is fed into the model. But in reality, the spread of the virus would not follow a specific trend, we are just able to replicate the previous data and also considering the demographics, living conditions of people, the possible rates or its increase could to a certain extend be predicted. The SIR model relies on patient population data and its changes with time. It takes into account three parameters, the people with possibility of infection(S), infected people from the pool of people who had probability of being infected(I), infected and later on recovered people(R). now using these three data a probability calculation is estimated that could show the probability of infection in that particular area and similar demographics. These data when correlated with the surge rate we are able to produce a satisfactory curve on the increase of patients in the upcoming months, hence the rate of spread of the disease can be easily predicted. Similarly, mathematical modelling using various other data could produce further accurate predictions and its combination would make the predictions even more reliable. SIR and SEIR AI models help in drawing various analytical graphs and study on the spread in various places and the similarities shown, we could predict the overall infectious nature of the virus across the globe and find reasons that deteriorate its spread which would facilitate in easier solution development.

As we have now seen the process of fighting this covid 19 pandemic, we could anticipate that the modelling of virus pandemic behaviours would be aided by artificial intelligence in the future. It could provide mathematically regulated estimate for the ideal lockdown period that is not too short to facilitate the pandemic to resurface nor too long to harm the economy and be driven into extensive rates of unemployment rates and recession. It could be ensured that the death toll, infected rate, the economic damage, the lockdown period and necessary resources be minimised. And the availability of relief facilities at the exact required places be maximised as we will be able to predict the places with possible extensive increase. The treatment costs would be reduced and would help us choose intelligently the right mixture of preventive strategies for its control.

Overall, the artificial intelligence system rather than just providing a mathematically driven statistical analysis, it simulates the entire pandemic conditions using the bulk of data using super computers to predict with more accuracy. With the bulk of intelligently derived fight back strategies AI could also be deployed in medicine R&D which would ensure faster medical support and hence a faster break from the pandemic. So, these AI models would ensure a completely systematic and strategized fight against any devasting life threatening situation that may arise in the near future and would ensure our readiness against facing the global misery.