mtjvankuik / FluPrediction

Prediction of flu in the UK using text mining on tweets

FluPrediction

A demo application using text mining on TwitterSearch feed data to display predictions on Flu in the UK

Start demo

1. Install Webstorm from IntelliJ. You can find the JavaScript IDE by navigating to the following URL: Webstorm A 30-day free trial is included.
2. Unzip the .zip project file to a known location.
3. Once Webstorm is successfuly installed, in the navigation bar located at the top select File -> Open..
4. Select the Project root folder and select Open.
5. Select the bin/www configuration and Run the project.
6. Go to your browser and navigate to localhost:3000
7. The Web app should be running and displayed in the browser.

Important to note

1. Pagination (e.g. retrieval of multiple pages of of tweets from twitter query) is turned off by default. You can turn this on by navigating to twitterSearchGet.js and uncomment line 78. //retrieveTweetsBatch(maxID,callback);
2. Only the first 100 most recent tweets that fulfill the query are used for prediction.
3. There is a limit to the amount of requests you are allowed to send to Twitter. This resolves quickly within a few minutes.
4. If you want to receive the latest tweets that have been generated in the meantime, run analysisAutoML.js. If you see tweets returned to the console, that means the latest tweets have been analysed, the classifier has predicted whether the retrieved tweets are actually flu and removed non-flu tweets, and a new data/cities.json file has been generated.
5. In the folder data/tweets.csv a list of 5000 tweets including locations are saved. These may be used for testing purposes.