The Class 12 Capstone Project
In Class 12 the AI project becomes a Capstone Project: a group project that follows the full Data Science Methodology, solves a problem tied to a Sustainable Development Goal, and is presented in a three-minute video. It is worth 25 of the subject's 100 marks. This page explains what is asked, and gives ten worked capstones to learn from.
These are sample capstone projects, written for reference and self-learning. Each one is a practical illustration of the Data Science Methodology — how a real problem is understood, how data is chosen, prepared and modelled, and how a model is validated, deployed and improved through feedback.
They are not meant to be submitted as they are. Most datasets are samples made by a program. Use these projects to understand the method; then, as a group, choose your own problem, collect your own data, and write your own documentation. The viva is based on your capstone, and it will show whose work it is.
1Where the marks are
| Part C (2026-27) | Marks |
|---|---|
| Capstone Projectcapstone | 15 |
| Project Documentationcapstone | 6 |
| Video (exactly 3 minutes)capstone | 4 |
| Practical File | 10 |
| Lab Test (Python and Orange Data Mining) | 10 |
| Viva Voce (Capstone Project + Practical File) | 5 |
| Total | 50 |
2The capstone guidelines
Minimum 3 and maximum 5 students per project.
The project must address one of the Sustainable Development Goals.
The capstone is completed in Class 12 together with its documentation.
Exactly three minutes, covering six named parts.
The video must have these six parts, in this order:
- Problem statement
- The SDG the project is aligned to
- The AI concepts, domains and algorithms used
- How the project works
- Conclusion
- Acknowledgement to the teacher
Every project here ends with a video script timed to exactly three minutes, with those six parts.
3The Data Science Methodology
The handbook teaches the methodology put forward by John Rollins at IBM: ten steps in five modules. Each capstone report here has one section for each step, in this order.
- 1. Business Understanding
- 2. Analytic Approach
- 3. Data Requirements
- 4. Data Collection
- 5. Data Understanding
- 6. Data Preparation
- 7. Modelling
- 8. Evaluation
- 9. Deployment
- 10. Feedback
The methodology is a cycle: feedback from deployment leads back to the model, and often back to the data. Every report says what feedback would change.
4Ten capstone projects
Each is aligned to an SDG, and between them they use every Class 12 unit that can be run on a school computer: regression and classification, computer vision, big data, neural networks and data storytelling. So far they draw 75 charts and carry 88 test cases, every one actually run.
A capstone in regression: predicting a plot's harvest from its crop, weather, soil and fertiliser — Linear Regression against a Random Forest, validated with K-fold, and checked again after a drought season.
The curriculum's own data-storytelling capstone: thirty-five years of enrolment, attendance and dropout, told on Freytag's pyramid — with the diagnostic analysis to back the story, and the honesty to say what the data cannot prove.
A classification capstone: Logistic Regression and a Decision Tree screen check-up patients for heart disease, the threshold is tuned for recall, and the follow-up feedback finds the model misses more women than men.
Forecasting a city's daily peak demand from weather and the calendar — Linear Regression, a Random Forest and a neural network, tested on a year they never saw, cross-validated through time, and caught drifting by the feedback step.
A field-kit test in the morning, a lab verdict two days later — this capstone predicts the verdict on the spot, and makes the Unit 1 topic of missing values the heart of the project by comparing four ways of handling them.
Computer vision and a neural network on real handwriting: a multi-layer perceptron learns to read digits, OpenCV cuts PIN codes off envelope images — and the feedback step shows why 98% per digit is not 98% per letter.
A rare-event capstone: a next-day flood warning from rain, river and dam data, where a model that never warns would score 95% — so it is judged on recall and false alarms, validated monsoon by monsoon, and replayed on a new monsoon as feedback.
A natural language processing capstone: citizens' complaints, typed in their own words, are cleaned, turned into numbers with TF-IDF and routed to the right department — and the ones the model is unsure of go to a human officer.
A big data capstone: a city's road sensors report every five minutes, without stopping. The data is read in chunks, summarised map-and-reduce style, cleaned piece by piece, and used to warn the control room fifteen minutes before a road jams.
An unsupervised learning capstone: K-means clustering finds groups of districts with similar health needs from eight survey indicators — with no labels to learn from — so that each group can get the programme it needs instead of one plan for all.