[2H-SHOULD] Generate confusion matrix + ML evidence figures

[YURDAKULOGLU]

Goal

Slide 6'daki 92.86% rakamını destekleyecek ek görsel kanıt üret. Soru gelirse "let me show you the confusion matrix" cevabı verilebilir.

Tasks

• tools/build_demo_extras.py'a confusion matrix figure ekle (3x3: sit/walk/run, gerçek prediction-vs-actual sayıları)
• Feature importance veya windows-over-time görseli (opsiyonel bonus)
• Çıktıları docs/presentation/figures/'a yaz
• Generator script'i güncelleyip commit et
• CI green olduğunu doğrula

Data source

docs/presentation/data/example_file_metrics.json — modelTrainingLabelCounts ve modelValidationAccuracy buradan.

Acceptance

• confusion_matrix.png üretilmiş ve commit'li
• generator script reproducible
• Stil rehberi uygulanmış (yellow/cyan/green palette, Courier New)

Estimated

30 min

[YURDAKULOGLU]

Confusion matrix + training label balance figures generated and pushed in c1b27b3:

• docs/presentation/figures/confusion_matrix.png — 3x3 grid, 13/14 correct, single walk->run boundary confusion
• docs/presentation/figures/training_label_balance.png — sit 27 / walk 22 / run 28

Generator updated: tools/build_demo_extras.py (run py -3 tools/build_demo_extras.py to regenerate).

Slide-ready, fits the deck palette. Use either as a backup answer to 'show me the per-class breakdown' or as additional asset for slide 6 expansion.