Learn Git and GitHub properly, clean a genuinely messy student dataset, work out which factors predict whether a student passes, and choose a model you can explain. Over seven days you'll build one public GitHub repository with a README and a notebook anyone can read, with no code run needed. Each day takes.
The week at a glance
Day 1 · Thu, Oct 22
Git and your first commit
Learn how Git tracks changes (60 min), then open your repo in a Codespace and make your first commit from the terminal (30 min). Finish both the exercise and the knowledge check. Stop your codespace when done. By tonight: a public repo with your first commit.
Day 2 · Fri, Oct 23
GitHub, README and Markdown
Learn the GitHub flow: branches, pull requests, issues (60 min). Then write your README skeleton with the one question you'll answer, create your folders, add the dataset, and merge a pull request (30 min). By tonight: your repo reads like a document, not a folder.
Day 3 · Sat, Oct 24
Clean the data
Study cleaning and exploring data (45 min). Then count the problems in the dataset before fixing any of them: wrong attendance values, negative study hours, missing values, text booleans. Fix each one and write down why. Create your pass/fail target (grade of 70 or more). By tonight: a cleaned dataset and a "Cleaning decisions" table in your README. Free alternative: Kaggle Learn's Pandas and Data Cleaning courses.
Day 4 · Sun, Oct 25
Exploration and feature importance
Learn permutation importance and charting (45 min). Then explore your data, build a correlation heatmap, and compare three ways of measuring which features matter. Run the comparison twice, with and without PreviousGrade, and note how the ranking changes. By tonight: you can say which factors matter and why the three methods disagree.
Day 5 · Mon, Oct 26
Picking the right model
Learn how models work, validation and overfitting (45 min). Then split your data, train at least two models (for example logistic regression and a decision tree), and compare their scores. Pick one and explain in your README why you chose it and what you gave up. By tonight: a chosen model with a reason that isn't "it scored highest".
Day 6 · Tue, Oct 27
Evaluation
Learn how to measure and diagnose a model (45 min). Then build a confusion matrix on your test data and write it out in plain sentences. Look at the cases your model got wrong and describe the pattern in one sentence. Say which mistake you'd rather make, and why. By tonight: you can explain what your model gets wrong in terms a teacher would understand.
Day 7 · Wed, Oct 28
Finish and share
Polish day, and a catch-up day if you fell behind. Finish your README with your cleaning decisions, findings, model choice and limitations. Read it as if you've never seen the project, then post your repo link to the community. By tonight: a repo a stranger could read and understand.