Lab: Build the ticket classifier baseline

Module 3 · Lab · About 45 minutes · Needs: lessons 1, 3 and 4 (cross-validation, the confusion matrix, pipelines and joblib)

What you are building

Priya wants tickets sorted into refund, shipping, warranty or other before a person reads them. By the end of this lab BrightCart has its first classifier: one scikit-learn pipeline that takes raw text and returns a category, scored honestly against a do-nothing baseline on the 40 labelled tickets, with a confusion matrix and an error table that say where it goes wrong. The fitted pipeline is saved to ticket_classifier.joblib with its scores inside, the way a service would load it. Module 4's lab refits the same pipeline as its baseline. Later modules compare against the numbers recorded here (0.55 against 0.25), not the file. No model key is needed; everything runs locally in Colab.