lillaszulyovszky/ironhack-final-project

A tool for managers/teachers to gauge student activity/engagement/mood based on Slack data

DA Bootcamp Slack Insights

Analyse student engagement/activity/mood with insights from your Slack Team's data.

Project Intro

Remote environments give very little chance for teachers to gain feedback and optimize their content towards better student performance. So I thought I could run some analysis and provide a tool for teachers to gauge student activity/engagement/mood and even predict what's the best way to get a reply (aka get help in this context) all based on Slack data. I only used data from public channels.

Notebooks

The notebooks in the code folder could be used to

• clean & wrangle json data to extract features into a dataframe
• run ML models for predictions
• run NLP preprocessing steps and sentiment analysis if you have your exported files at in hand.

If you don't know how to export, read up here> https://slack.com/intl/en-hu/help/articles/201658943-Export-your-workspace-data

The notebooks are fully annotated and include all the modules which needs to be imported to make the code work.

Future scope

Enhanced scope for this will include

• dynamically pulling data from Slack API (so insights can be drawn anytime/anywhere)
• loading publicly shared files into a google sheet for the whole cohort
• working as a Slack App, providing insights for teachers on the Slack space they are in.

Couple of insights to share

• distribution of messages between participants (students and teachers) in the top 5 most used channels

• distribution of messages during the day in the homework(lab) help channel

• identifying influencers by messages sent

• mood changes during the bootcamp in April on the homework(lab)help channel

• top perfoming emojis

• top 15 important features to look out for when you're looking for a reply

• most used words in the bootcamp

Methods used

• Loading JSON files: creating a function to load each file into a dataframe
  
• Data cleaning & wrangling in Python: transforming data set to help visualise insights, feature engineering
  
• Prepocessing: 2 methods (Normalizer, Dummies) for Predictions and several NLP preprocessing steps like removing punctuations, emojis, links, stemming
  
• Machine Learning Models
  linear regression, logistics regression, random forest, random forest classification
• Natural Language Processing
  wordcloud, VADER analyis, Sentiment analyis
  

Thank you for reading!
Any questions, hit me up at [email protected]