Predicting Workshop Attendance is a practical course for learners who already know basic pandas and regression. A community studio needs an attendance estimate seven days before each workshop so it can prepare materials. You will build a repeatable analysis that uses only information available at the appropriate time, measures real prediction errors and compares a small model with a simple baseline.
Two Guided Labs introduce the prediction deadline, a training freeze, date-based evaluation, training-only missing-value handling and a short regression Pipeline. One independent Challenge applies the same mechanisms to another season. Every unit starts with its own prepared data; files and processes do not carry between labs. The total 165-minute duration is an author estimate including reading and ordinary corrections, not measured beginner timing.
What You Will Learn
- Distinguish seven-day booking snapshots from final bookings and post-event scans.
- Freeze training before later prediction dates and reserve later labeled workshops for evaluation.
- Fit missing-value preprocessing on training rows and reuse it without learning from the holdout.
- Compare continuous regression predictions with a training-median baseline using MAE in people.
- Associate actual values, predictions and plotted errors by workshop ID.
- Export predictions for future workshops without pretending they have known outcomes.
Who This Course Is For
This course is for learners ready to turn basic regression into a small, honest planning workflow.
Prerequisites: Python functions, terminal input, pandas table selection and joins, basic Scikit-learn regression, train/test separation and introductory Matplotlib. Choosing Opening Hours from Data is recommended practice for interpreting data and its limits.
Learning environment: A browser-based WebIDE contains the original synthetic CSV data, Python, pandas, NumPy, Scikit-learn and Matplotlib. No local installation or external dataset account is needed. Programs, outputs and interface labels remain English.
Frequently Asked Questions
Must the model beat the baseline?
No. A correct comparison can show that the model adds no value. The goal is a truthful evaluation, not a required score.
Will I tune many models?
No. You use a small mean-imputation and linear-regression workflow. The main decisions concern what is known, when it is known and what the errors mean.
Are these production forecasts?
No. The small datasets are original synthetic teaching material. They demonstrate a reproducible workflow without establishing reliability for a real venue.
