/nupic.hangouts

An example of using NuPIC to detect anomalies in Google Hangouts conversations.

Primary LanguagePython

nupic.hangouts

An example of using NuPIC to detect anomalies in Google Hangouts conversations.

Usage

  1. Install dependencies.

    Follow installation instructions for Unicorn.

    Install matplotlib.

  2. Install nupic.hangouts.

    python setup.py install --user

  3. Get your data.

    Visit Google Takeout, select only the "Hangouts" option, and download. Extract the Hangouts.json file from the downloaded data.

  4. Process your data.

    python -m nupic_hangouts.scripts.process PATH/TO/Hangouts.json data

  5. Run a conversation through a NuPIC model and plot results.

    Pick a conversation in the data directory (one of the .csv files). The number in the name of the .csv file indicates how many total messages are in the conversation.

    ./model.py PATH/TO/CONVERSATION.csv output

Example output

Example Output The red triangles are points of high anomaly.