Using information from occurrences of covid19 made available by NASA, national data from the existing public infrastructure, symptoms and chronic illness data from patients collected by the technology of the voice assistant (June Tracker), it is possible to identify the location of people with symptoms of covid19 and categorize them by risk level, we can provide information on the evolution of the disease (cases and deaths) and the level of occupation of hospital infrastructures using machine learning algorithms. This intelligence allows identifying high-level locations, helping to maintain social distancing, and the health infrastructure available.
Documents below:
Covid Tracker: https://github.com/dlsouza42/CovidTracker
June Tracker: https://github.com/thiagorbernardo/nasa-hacka
June Tracker Demonstration: https://youtu.be/t2qYNk9TSeg
Additional Information:https://drive.google.com/drive/folders/16fKEMyhG4TUWVbkFyUI8Rz5frvfMbJ0c?usp=sharing
Our team developed a co-creation / brainstorm process and compiled the ideas on the FunRetro platform. During co-creation, the following topics were considered: Public, Value Proposition, Competitive Advantage, Channels, Problem, Solution, Key Metrics, Cost Structure, Sources of Revenue. After discussion by the team, the following problems were verified:
1 - There is no priority organization of telemedicine consulting
2 - lack of follow-up and guidance of the population that has symptoms of COVID and uses the public and private health system
3 - Lack of daily clinical follow-up for cases that report symptoms and do not perform the COVID exam
4- Hospitals have difficulty planning their service structure, as they do not have the capacity to analyze the occurrence of disease cases at the municipal level
5 - Several patients who have other symptoms of serious illnesses fail to go to hospitals due to an existing health calamity In order to create a winning solution, the need was identified to create a personal assistant to communicate with people who show symptoms of the disease and collect local data from contagion clusters and carry out the increase of cases, and use of the health system used by non-parametric algorithms.
What motivated us to participate in the challenge was the possibility of helping the population and health professionals, within our area of expertise, which is focused on technology. The idea of being able to keep people who do not need care at home with guidelines within international standards and those who need to contact health professionals by telemedicine and refer hospitals with fewer occupations to those who really need them, made us believe that we all have a contribution in this moment of so many challenges for the nation.
With the user symptoms and comorbidities picked by June Tracker (Voice Assistant) and the NASA Compendium (https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge/data#) made with open data of COVID-19 researches, we estimate which factors are more likely to bring someone to die, and estimate the person class of risk. Municipal case forecast data is taken from the connection (https://sedac.ciesin.columbia.edu/mapping/popest/covid-19/) provided by NASA
June Tracker is a action made for Google Actions, it is used DialogFlow agent, a NLP (Natural Language Processing) platform where we can build conversational journey with a Voice User Interface (VUI) to guide the user for a better advice. The agent uses Firebase functions (hosted on Google Cloud) and is used Javascript to program the function.
The data to the forecast of municipal cases come from the connection provided by NASA. For the estimation, we used the Facebook’s Prophet algorithm in R language.
To store user data (symptoms and comorbidities) we choose MongoDB, a No-SQL Database, where you can store JSON based documents.
As a team, we take the opportunity to share how we can create with agility and improve knowledge in different areas such as: Data manipulation, VUI, UX, UI, Prototyping, Project management. Throughout the process, the environment was maintained in cooperation and synergy in order to stay focused and, at times, abstract the pandemic situation, given the energy expended throughout the process. And in the end, but not less we created a complete demo to be presented to the challenge.
Two problems were found with the tableau: first, the tableau did not connect with Mongo DB and there was instability in the publication of the Tableau Online dashboard.
https://sedac.ciesin.columbia.edu/mapping/popest/covid-19/
https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge/data#