Exercises

Chapter 9 of Module 17 — Django for AI Web Apps. Do these against the running example-app/.

Each exercise has a worked solution — try first, then peek.

1. Add a field

Add region (a ChoiceField: north/south/east/west) to ChurnForm, the Prediction model, and the admin list_display. Re-run makemigrations + migrate.

💡 Solution

Add region = forms.ChoiceField(choices=[("north","North"),...]) to ChurnForm; add region = models.CharField(max_length=10) to Prediction; add "region" to PredictionAdmin.list_display; then python manage.py makemigrations scoring && python manage.py migrate. (The scorer can ignore region or add a small per-region offset.)

2. Filter the history

The shipped /history/ page shows every prediction. Add a ?contract= filter: /history/?contract=month-to-month should show only those rows, pagination intact, and the page should offer the three contract types as links.

💡 Solution

# views.py — filter BEFORE paginating, so page counts stay correct
@login_required
def history(request):
    rows = Prediction.objects.all()
    contract = request.GET.get("contract")
    if contract:
        rows = rows.filter(contract=contract)
    paginator = Paginator(rows, per_page=10)
    page_obj = paginator.get_page(request.GET.get("page"))
    return render(request, "scoring/history.html",
                  {"page_obj": page_obj, "contract": contract})

Template: three links like <a href="?contract=month-to-month">month-to-month</a> (plus one plain ? link for “all”), and carry the filter through the pager links: ?page={{ page_obj.next_page_number }}&contract={{ contract }}. Querysets are lazy, so stacking .filter() before the paginator costs one SQL query either way.

3. Validate the API

Make api_score return HTTP 400 with a helpful message when a field is missing or non-numeric (instead of a 500).

💡 Solution

Wrap the parse in try/except (KeyError, ValueError, TypeError) as exc: and return JsonResponse({"error": str(exc)}, status=400). (The shipped api_score already does this — read it, then make the error messages friendlier.)

4. Test the threshold

Write a test asserting that a month-to-month customer with high charges and many tickets scores higher than a two-year customer with the opposite profile.

💡 Solution

def test_contract_matters(self):
    high = churn_probability(tenure_months=2, monthly_charges=110, support_tickets=6, contract="month-to-month")
    low  = churn_probability(tenure_months=40, monthly_charges=40, support_tickets=0, contract="two-year")
    self.assertGreater(high, low)

5. (Stretch) Swap in a real model

Train a small scikit-learn classifier on the synthetic churn data from NB 17 (05_machine_learning/17_sklearn_basics.ipynb), joblib.dump it, load it once in ScoringConfig.ready(), and call it from churn_probability. The views, forms, templates, and admin stay untouched — proof the seams are in the right place. (Stuck? Chapter 7 now walks the full pattern — attempt it first, then compare.)