We started our project with the intent to help prevent the spread of the Coronavirus
What it does
Predicts the path of a healthcare outbreak. We have special; transformation tools that can take WHO/Open source data and plot the spread. We can predict how far it can go. It may not be very accurate. We have universal api which can be used by anyone who wants to share their data. All covid Survivors are Welcome.
How we built it
python, tableau, r studio ,Aws kinesis,
Challenges we ran into
While cleaning the data set we ran into computing errors. We fixed them by editing the data set and clearing NANs.
Accomplishments that we're proud of
A working model
What we learned
How to create plots and predictive analysis that can hopefully be applied to future outbreaks
What's next for The Trackers
Applying this tool to track the next pandemics with the intent of preventing fake news. We hope to customize our tool to the next outbreak in order to save at risk groups.
our sample code
title: "SamplePredict" author: "Sneha Gondane" date: "2/29/2020" output: html_document: default pdf_document: default
require(tidyverse) data <- read_csv('C://Users/Rahul/Documents/SunFlowerHack2020/sample3.csv') data require(lattice)
install.packages('leaps') require(leaps) lps=regsubsets(type~Lat+Long+cases ,data=data, method='backward') plot(lps,scale='adjr2')
#To reproducd data from a consistent starting point using seed value require(tidyverse) set.seed(1234) #Select training data (using 60 / 40 ratio) training.data=data %>% sample_frac(size=.6) #training.data #Select testing data testing.data=data %>% anti_join(training.data,by='ID') #testing.data training.input=training.data%>%select(-type, -ID) #training.input testing.input=testing.data %>% select(-type, -ID) #testing.input training.label=training.data$type #training.label testing.label=testing.data$type #testing.label
require(class) predictions=knn(train=training.input, test=testing.input, cl=training.label, k=25) data.frame(predictions,testing.label) summary(data.frame(predictions,testing.label)) accuracy = sum(predictions==testing.label) / length(predictions) accuracy