Machine Learning · 7,000 participants · synthetic data
Social Media & Mental Health
Analyze 7,000 social media usage profiles to predict mental wellbeing. Three progressive notebooks: data exploration, Random Forest classification, then regression to predict the anxiety score.
PythonPandasNumPyMatplotlibSeabornscikit-learn
Free · No account·Intermediate·~2h30
What you'll build
- ✓A structured exploration of digital habits, sleep and wellbeing indicators
- ✓A dataset prepared for machine learning despite numerical, categorical and ordinal variables
- ✓A classification problem evaluated beyond accuracy while accounting for imbalanced groups
- ✓A regression problem compared with a baseline to judge whether the model adds useful information
Lab walkthrough (3 notebooks)
01
Understand the population
Inspect distributions, available categories and the quality of the information.
02
Transform data without losing meaning
Handle missing values and represent each kind of variable appropriately.
03
Explore relationships
Compare groups and use correlations as clues without mistaking them for causal links.
04
Classify wellbeing levels
Build a Random Forest, control its complexity and study its errors despite class imbalance.
05
Predict an anxiety score
Establish a baseline, train a regression model and compare predictions with observed values.
Skills covered
- · Heterogeneous data preparation
- · Encoding suited to variable types
- · Careful correlation analysis
- · Classification with imbalanced classes
- · Regression and baseline comparison
- · Feature importance interpretation
Prerequisites
- · Python basics (functions, lists, loops)
- · Basic Pandas (DataFrame, columns)
- · No installation : online lab