Predictive statistical research model for COVID-19

Igg Table(Antibodies)
SQLite database browser > Igg Table(Antibodies)
SQLite database browser > General Symptoms Table
General Symptoms Table
R Studio > piece of code
Neural Network Prediction Model

Inspiration

The proposal is a prototype called "Predictive statistical research model for COVID-19" adressed to the Challenge> Health>Research. Our NGO in called ASOCIAȚIA LUMEA LUI ALEXANDRU/ALEXANDRU'S WORLD ASSOCIATION and our objectives are dedicated to pacients with: Hunter Syndrome, tracheotomy, disabled in the fields of: medical, educational, social and research purposes.

Motivation

Rare disease are already a challenge and the new Coronavirus is adding more pressure.

What it does

Our solution creates a statistical predictive model, so the researchers can choose from multiple factors that can define a certain type of reaction, severity, treatment, demographic indicators(ex. the density of infected people in certain area), medical indicators(ex.Igg,IgM,antibodies etc.) and also other type of problems like: diabetics, obesity or heart issues. The outcome is prediction per patient, it can help model a certain type of patient profile.

How I built it

The proposed prototype has demonstrative data, and it can be extended and added more records and fields.We assume that in short time there will be a great amount of data linked with COVID-19 disease and a very good tool for getting some conclusions or patterns would be a model that can be customized in the sens that the physicians or the researchers to choose the fields that have the higher importance.

Technical data

We used: SQLite database, R for statistics, Shiny for interface and reactiveness and Neural networks for obtaining the predictive model. The outcome is a prediction of a risk to develop COVID-19, by taking into consideration as many factors as we want. We also assume that we have test results and that way our model can be validated.

Challenges I ran into

For now the data is only for demonstrative purposes and include some general symptoms and something about antibodies and we think that with bigger amount of real information the outcomes can be more spectacular.