Inspiration
Diseases like ebola gain attention after huge damage is done. Our idea would help not only the scientists, but also its day to day users to spot the changes in trends so that an action could be taken as soon as possible.
What it does
It searches for disease related keywords on social media (currently twitter) to get the results using their API (Using Twitter API). A series of algorithms involving classifiers segregates useful tweets that represent diseases of people with an accuracy of 87%. The user of the data is then identified into one of the 15 categories (like Parent, insomniac, alcoholic, etc) by analyzing 200 tweets for each eligible user. Using the analysis, the app displays disease count for the number of users, user categories for each disease and the general trend of the disease. The final result (visualization and trends) is then displayed on our website and the IOS App.
How we built it
we used django and python to create classifiers and wrote data mining algorithms which interact with our website via apis we setup used django. The classifiers were original to profile user and remove fake tweets were original contribution.
Challenges we ran into
It was a huge challenge to figure out efficient algorithms and also required a lot of brainstorming to bring the efficiency to 85%. The data collected from Twitter itself was huge. Due to time restriction, we could only run the script to collect data over 3 hours from 5 am to 7:45am and were able to handle 3000 tweets and store about 200 valid tweets with geolocations on most of them. The time constraint definitely made things a lot harder especially when there was so much to do.
Accomplishments that we're proud of
The efficiency of our algorithms and the visualization is definitely the highlight of the application. Apart from that, the application can handle Huge data influx as some of its modules use Amazons s3 and dynamodb (currently, only for the community in IOS app). The IOS application is also able to authenticate and authorize users using Facebook login and Amazon AWS cognito. The identity pool for the app is highly scalable and allows only authenticated users to submit data but is open for all if they want to view the visualization results.
What we learned
Data mining algorithms like apriori, fpgrowth and some fuzzy logic. IOS App development with objective-C and most importantly, the ability to work under pressure.
What's next for DTRAP
The next steps for DTRAP include adding the support of other social media platforms like Facebook and Reddit. Apart from that, the iOS application created for the disease would also take into consideration the community of DTRAP users where a user would be able to report his disease/symptoms that could be marked on the Map. Apart from that, push notifications can be sent to the users when they are about to enter an area that is marked red (possibility of an outbreak of a contagious disease).
Log in or sign up for Devpost to join the conversation.