Artificial Intelligence framework considered for Covid forecasting

2 November 2021 12:00 am Views - 553

 AI-based solutions help monitor human behaviour in ‘super spreader events’ such as entertainment, ceremonial and social events that gather people in large numbers to a selected location and identify how the disease will spread from such events

 

A research group from the University of Peradeniya is using Artificial Intelligence (AI) frameworks for ‘threat assessment and containment for COVID-19 ‘and future epidemics while mitigating the socio-economic impact on identified vulnerable groups. The main aim is to detect, model and predict the impact on identified vulnerable groups: women, children, and underprivileged people, under COVID-19 pandemic containment strategies and developing AI-based solutions to mitigate the future spread of COVID-19 and future pandemics. This research group was able to develop AI-based models to address several issues faced by the nation, said Prof. Janaka Ekanayake Head of the Electronics Division of the Faculty of Engineering of the University of Peradeniya and the Principal Investigator of the Project.


In addition “ a Covid vision CCTV “, model has also been developed to detect the areas where people do not follow health guidelines like social distancing and mask-wearing using computer vision techniques that were trained by the data collected from CCTV footage. This system consists of a framework to unify different computer vision algorithms, a graph-based representation to store the information extracted from CCTV footage, and a formula to holistically interpret the graph and quantify the threat level of a given scene. This enables us to visually identify the impact of human behaviour on the disease spread through video surveillance and implement regulations to minimize the threat.


To begin with Prof. Ekanayake said that it is appropriate to discuss the latest contribution on forecasting daily new Covid-19 cases and deaths in Sri Lanka. “As opposed to existing models that predict the long term trends, they use a Deep Learning based AI model to predict the new Covid-19 cases and the expected death toll 10 days ahead, using past data. “This model was trained using historical data from a diverse set of states and countries around the globe like Italy, Japan, Nigeria, the Texas state of the US. These countries faced varied circumstances due to the pandemic which will enable the trained ‘model’ to better learn most situations that might occur due to the pandemic irrespective of the country, making it a more general predictor with the capability to predict the pandemic situation of many countries including Sri lanka. The prediction results for Sri lanka are available on their official website [University of Peradeniya - COVID Research Group (pdn.ac.lk)] that illustrates the prediction of new Covid-19 cases and deaths. The results from the proposed model can assist government policymakers to make rapid and proactive decisions in the short term,” said Prof. Ekanayake.

 


Funding for the programme 


The research team is led by Prof. Ekanayake. Prof. Vijitha Herath, Dr. Roshan Godaliyadda and Dr. Parakrama Ekanayake from the faculty of engineering investigate the Data analytics and Artificial Intelligence front while Prof. Mallika Pinnawala and Prof. Sakunthala Ekanayake from the University of Peradeniya and Dr. Ganga Thilakarathne from Institute of Policy Studies led the research on the social science front. Further, Prof. Samath Dharmarathne Faculty of Medicine and Dr. Mahen Kothalawala guided the team with their expertise on medicine. The research assistants in the group are Gayanthi Anuradha, Yasiru Ranasinghe, Harshana Weligampola, Erandi Attygalla, Gihan Jayathilake, Umar Marikkar, Jameel Hassan, Suren Sriharan  and Rumalie Perera. The research is funded by International Development Research Centre (IDRC), Canada, and the Swedish International Development Cooperation Agency (Sida) through the Global South AI4COVID programme. This research is a collaboration between the University of Peradeniya, Institute of Policy Studies, Sri Lanka and the University of Tenaga, Malaysia.


This model was trained using historical data from a diverse set of states and countries around the globe like Italy, Japan, Nigeria, the Texas state of the US. These countries faced varied circumstances due to the pandemic which will enable the trained ‘model’ to better learn most situations that might occur due to the pandemic irrespective of the country, making it a more general predictor.
- Prof. Janaka Ekanayake Head of the Electronics Division of the Faculty of Engineering of the University of Peradeniya

 



 Further, their research also focuses on building a computer model that tries to understand how an individual’s action and behaviour at the micro level affects the pandemic at the macro level. This model is capable of emulating human behaviours that resemble the daily routine of employees from the manufacturing, commercial, transport, health, and education sectors and how their day-to-day interactions affect disease transmission. As the country gradually reopens and relaxes restrictions it would be prudent for us to better analyze the impact of such human actions and how they will ultimately affect the disease spread. A special feature of this model is that it examines the role public transport plays in disease transmission. Additionally, it simulates human behaviour in “super spreader events” such as entertainment, ceremonial and social events that gather people in large numbers to a selected location to identify how the disease will spread from such events. Also, the research group is hoping to develop the simulator to identify the impact of social distancing, different vaccination and containment strategies. Furthermore, they have developed a questionnaire for a large household survey to assess the socio-economic impacts of identified vulnerable groups especially on women, children and underprivileged groups. Basically, this survey will capture impacts on social, economical, health and mobility aspects. The aim is to survey more than 3000 households covering all districts in the country. This survey will be administered in the near future adhering to strict health guidelines. The outcomes of this survey will be used to tune the simulator to properly model the human impact of the pandemic.


Further, their research also focuses on building a computer model that tries to understand how an individual’s action and behaviour at the micro level affects the pandemic at the macro level. This model is capable of emulating human behaviours that resemble the daily routine of employees from the manufacturing, commercial, transport, health, and education sectors and how their day-to-day interactions affect disease transmission. As the country gradually reopens and relaxes restrictions it would be prudent for us to better analyze the impact of such human actions and how they will ultimately affect the disease spread. A special feature of this model is that it examines the role public transport plays in disease transmission. Additionally, it simulates human behaviour in “super spreader events” such as entertainment, ceremonial and social events that gather people in large numbers to a selected location to identify how the disease will spread from such events. Also, the research group is hoping to develop the simulator to identify the impact of social distancing, different vaccination and containment strategies. Furthermore, they have developed a questionnaire for a large household survey to assess the socio-economic impacts of identified vulnerable groups especially on women, children and underprivileged groups. Basically, this survey will capture impacts on social, economical, health and mobility aspects. The aim is to survey more than 3000 households covering all districts in the country. This survey will be administered in the near future adhering to strict health guidelines. The outcomes of this survey will be used to tune the simulator to properly model the human impact of the pandemic.