This project aims to identify diabetes distress patterns based on social media data using artificial intelligence methods. We are working with twitter data.

In the following a small overview over the directories:

• Visualisation_US_map: D3 visualisation of tweets occurrence over the USA after our geolocation algorithm
• Tweets_extraction_Twitter : Extractions of tweets via Twitter API
• WordEmbeddings : Calculating word embeddings (Word2Vec or FastText) via the gensim package
• data : Trained models (not up-to-date)
• db : Algorithms to filter tweets, remove duplicates (from chatbots) , clean database , ..
• files : Only list with keywords to extract tweets
• jupyter_notebooks : To experiment
• preprocess : Functions to preprocess tweets and textual data in general
• readWrite : Read & Write files (parquet, csv, text)
• tests : (not up-to-date)
• topicModel : Extract topics with LDA method
• training : Train classifiers for filtering or predicting
• utils : utility functions

More detailed information about the programs and algorithms used, you will find in the corresponding folders.

Check the development branch 'devAA' for current programs.

• Python (version >= 3.5.5)