CORPORATE/ PM ROLE/ SOLUTION
COVID 19 and the role of ML
Suggestion on how ML could play a role in COVID detection in Mar — 2020 (Copy of Old Article)
I recently had a discussion with one of my friend who was trying to work on the datasets present on Kaggle related to COVID 19. This data set is getting updated every 24 hours and the problem that they are looking into is to predict the number of cases that are going to happen based on different parameters, one being the Region.
Another piece of information came in from an article which described different types of victims of the Corona Virus and why is it hard for any government to keep a check on this number without a lockdown. The article segregated the victims into three categories, the First being Case A, which is the positive case of Corona. He travels to different places until finally found as a positive case. He met some of his known people who are going to become Case C victims. The government can track both cases A and case C victims. However, he might have met a few unknown on the way say in Public places like metro, etc they are the people who can become a victim. Let’s categorize them as Case B victims. These Case B victims in a span of days will meet a few others. We can categorise them as Case D victims. These Case B and D victims are the ones hard to find. Thus LockDown is going to put a hold on this number.
This seems to be a logical move, but in a country like India can we only depend on the people to stay inside so that a person with the symptom is not affecting others. More than COVID-19 is the problem of food especially for poor people who have documents from other states and working in another. A case arose today where thousands of people started to march towards UP/ Bihar walking. They were so close to each other this was going to increase the chain of cases furthermore.
I have also heard from many that the COVID is spread to riches and the poor are unaffected from it yet, as this disease is not something that can evolve here. It can only happen to those who came in contact with foreigners or had foreign trips. They forgot that many people work in the houses of such people and they even go to market touching vegetables (even when talking, a small amount of saliva may be spit from the mouth) and other items and these items can be touched by any other person be it poor or middle class. Hence, even in the Lockdown, the surrounding could be full of the virus which we are unaware of.
We have discussed a lot what are the challenges and most of us till now are very much aware of the situation and challenges. I chose to write down not just on issue, however on how can ML help in identifying a small percentage of Case B and Case D victims.
In India, we have Aadhar cards for the majority of our population. Aadhar card database has the most crucial information about every individual ie our picture and basic data. When a Case A victim is identified, he will be entering the nation through some airport. The airport will have a camera at every corner, thus through the facial recognition of all the user Case A victims came in contact a chain could be started. The facial recognition could be done with photos present in the Aadhar database (Other DBs could also be integrated with).
The detail of all the people he came in contact with could be then identified from different cameras present in the societies, metros, ATMs etc. This chain could be increasing with identified people. The data of these identified could be decreased by using ML where, using different parameters such as contact distance, type of contact, age, region, the temperature of the area during contact, avg temperature of the person living in, type of work, avg working hours, gender etc we could get an estimate on who is more capable of getting infectious. These can be compared with data of new cases that are coming in and were already the part of the chain.
Thus using facial recognition on different recordings and live cameras we can start making a list of affected and a testing team can then be sent to these people on the basis of severity the ML model suggests. These test results need to also be trained so that more accurate suspected chains could be discovered within less time.