AI Model Development | CS PPT +Worksheets +Activity + Mcq Quiz + Answers_Marking Scheme

1. Lecture PPT (14 slides, dark navy/purple + neural-net decoration)
   Title → Objectives → AI Lifecycle (7-stage pipeline) → Data Preparation (6 techniques: cleaning, feature engineering, encoding, scaling, imbalance, splitting) → Model Training (training loop, loss functions, optimisers) → Overfitting & Underfitting (3-column comparison + causes/fixes) → Regularisation Techniques (L1, L2, Dropout, Early Stopping, Cross-Val, Augmentation) → Neural Network Architecture (diagram + forward/backprop equations) → Evaluation Metrics (confusion matrix, Accuracy, Precision, Recall, F1, ROC-AUC) → Deployment & MLOps → Worked Example: CNN Image Classifier (full Python code annotated) → Worked Example: Credit Card Fraud Detection → Practice Questions → Summary

2. Worksheet (70 marks)
   Section A: Lifecycle stages table + iterative explanation · Section B: Data cleaning tasks, Min-Max vs Z-score, imbalanced data, data leakage · Section C: Gradient descent, loss functions table, learning rate problem, mini-batch comparison · Section D: Overfitting diagnosis + solutions, neural network architecture, rare disease metric choice, confusion matrix calculations · Section E: Deployment considerations, drift types, AI credit decisions ethics

3. Class Activity — “Model Builders”
   4 real ML problems (Healthcare Sepsis / Education Dropout / E-Commerce CLV / Cybersecurity Intrusion) · Teams plan full pipeline: data audit → model design → evaluation strategy → ethics & deployment → board pitch · Class debate on autonomous AI decisions and accountability

4. MCQ Quiz (25 questions, 30 mins)
   5 sections: Lifecycle & Data Prep · Model Training · Overfitting & Regularisation · Neural Nets & Evaluation · Deployment & Applied

5. Answers & Marking Scheme (Teacher Copy)
   Full model answers with tick-point marking, model code for confusion matrix calculations, examiner notes, MCQ key, and grade boundaries (combined 95 marks)