CBC Intelligent Tutoring System for marginalized learners in Kenya
# π Elimu AI β CBC Intelligent Tutoring System
> **AI-powered offline-ready tutor for marginalized learners in Kenya**
> University of Eastern Africa, Baraton | AI Group 19
> Delfine Achieng Β· Geoffrey Njenga Β· Chepchumba Faith
---
## π Project Overview
Elimu AI addresses the educational gap in arid, remote, and conflict-affected regions of Kenya (Turkana, Garissa, Mandera, Marsabit Counties). It is an offline-capable AI tutoring system that:
- Uses a **Random Forest Classifier** (supervised ML) to classify student performance
- Covers **Mathematics, English, and Kiswahili** β aligned to the Kenyan CBC curriculum
- Works **without internet** via a locally embedded model (offline-first architecture)
- Is hosted on **Hugging Face Spaces** (free cloud) for when connectivity is available
- Runs on **solar-powered, low-cost devices** (no heavy compute required)
---
## π€ Machine Learning Algorithm
**Algorithm: Random Forest Classifier** *(Breiman, Leo. 2001)*
| Metric | Value |
|---|---|
| Test Accuracy | **100.00%** |
| Validation Accuracy | **99.92%** |
| 5-Fold CV Mean | **99.96% Β± 0.03%** |
| Training Samples | 5,600 |
| Performance Classes | Beginner Β· Intermediate Β· Advanced |
### Features Used
| Feature | Description |
|---|---|
| `grade` | Student's CBC grade level (1β6) |
| `subject_enc` | Subject (Math / English / Kiswahili) |
| `topic_enc` | Curriculum topic (encoded) |
| `difficulty_enc` | Easy / Medium / Hard |
| `num_questions` | Questions attempted |
| `correct_answers` | Number answered correctly |
| `incorrect_answers` | Number answered incorrectly |
| `accuracy_rate` | Proportion correct |
| `total_score` | Score out of 100 |
| `avg_time_seconds` | Average time per question |
| `score_per_question` | Score efficiency ratio |
| `efficiency` | Accuracy / time ratio |
### Data Split
- **70%** Training (5,600 records)
- **15%** Validation (1,200 records)
- **15%** Test (1,200 records)
---
## ποΈ Project Structure
```
elimu-ai/
βββ data/
β βββ β¦