Pranav activity model

[Siphon03]

Description

This PR adds activity recognition and fall detection models using both mHealth and SisFall datasets.

The implementation includes:

• Data preprocessing and cleaning of sensor datasets
• Training of three machine learning models:
  • Random Forest
  • Support Vector Machine (SVM)
  • K-Nearest Neighbors (KNN)
• Model evaluation using accuracy, classification report, and confusion matrix

Random Forest achieved the best performance (~85% accuracy on SisFall dataset).

---

Todos

✔ Tested and working locally
✔ Code follows project guidelines
✔ Self-review completed
✔ Changes documented

---

How to test

1. Navigate to AI_models_pranav folder
2. Run:
3. Output will display:

• Model accuracy
• Classification report
• Confusion matrix visualization

---

Screenshots

Included in this PR:

• Accuracy comparison graph
• Confusion matrix
• Feature importance graph

(All images are available inside the AI_models_pranav folder)

---

Known Issues

• Large dataset size required sampling for faster training
• Some imbalance in fall vs normal activity classes