Why
Social media group, such as Messenger Group, is a rich ground for data mining to extract information about users interaction. Here I demonstrate some basic data exploration for a small group (The data is actually from a group of my friends, with names anonimized). For a large group with more data, a lot could be done, for example, finding cluster of users in more features space.
The Jupyter Notebook
Load facebook data for a group chat (json files) into data frame, properly decode unicode Vietnamese, and explore the data (number of messages daily, monthly by senders, plot frequently used words and emoticons).
Frequency of interaction
Number of messages, reactions by user, monthly average.
Interaction Matrix
From the messages, you can extract an interaction matrix between users: Who is mostlikely to respond to whom in the group.