Skip to content

project_report.md

Latest commit

 

History

History
257 lines (180 loc) · 9.56 KB

project_report.md

File metadata and controls

257 lines (180 loc) · 9.56 KB

🧠 TaleemAI: Personalized Adaptive Learning + Teacher Support System

Developed by: Syed Mushahid Ali Kazmi Supported by: Muhammad Abdullah Repository: TaleemAI on GitHub License: MIT Version: 1.0 Language Focus: Urdu / Regional Languages (Accessibility Module) Platform: Google Colab + GitHub Actions (Automated ML Workflow)

Python License GitHub Workflow Status Platform AI

📘 Abstract

TaleemAI is a personalized adaptive learning and teacher-support system designed to enhance education through data-driven insights, accessibility, and automation.

It leverages Machine Learning (ML) and Natural Language Processing (NLP) to:

  • Analyze student performance
  • Predict outcomes
  • Support teachers with actionable insights
  • Provide adaptive learning in Urdu and regional languages

Modules include classification, forecasting, and language translation, combined with automated reporting and GitHub CI/CD integration.

🎯 Project Objectives

# Objective Description
1 Personalized Learning Deliver customized insights based on student performance.
2 Teacher Assistance Provide data-driven suggestions to educators.
3 Accessibility Translate educational content into Urdu and local languages.
4 Automation Automatically train, evaluate, and update models using GitHub Actions.
5 Reporting Generate professional PDF reports and visual insights.

🧩 Problem Statement

Many educational systems lack:

  • Personalized student analytics
  • Teacher-oriented data insights
  • Regional language support

TaleemAI addresses these challenges by:

  • Predicting student performance trends
  • Providing actionable teacher analytics
  • Offering Urdu and regional language translations
  • Automating the end-to-end pipeline

📊 Dataset Information

Dataset Source: Kaggle - Students Performance in Exams

Description: Demographic info, parental education, test preparation, and exam scores (Math, Reading, Writing).

Feature Type Description
Gender Categorical Male / Female
Race/Ethnicity Categorical Student's ethnic group
Parental Education Level Categorical Highest education of parent
Lunch Categorical Free / Standard
Test Preparation Course Categorical Completed / Not Completed
Math Score Numeric Math performance
Reading Score Numeric Reading performance
Writing Score Numeric Writing performance

🧠 Methodology

TaleemAI Pipeline: Five stages

1️⃣ Data Preprocessing

  • Load dataset via Kaggle API or CSV upload
  • Clean missing values and normalize numeric data
  • Encode categorical features (LabelEncoder / OneHotEncoder)

2️⃣ Model Training (Classification)

  • Target: Pass/Fail based on average score
  • Algorithms: Logistic Regression, Random Forest Classifier, Naive Bayes (NLP tasks)
  • Metrics: Accuracy, Precision, Recall, F1-Score, Confusion Matrix

3️⃣ Model Training (Forecasting)

  • Goal: Predict future scores
  • Algorithm: Linear Regression
  • Output: forecast_results.csv

4️⃣ NLP & Translation Module

  • Tools: TextBlob, Googletrans
  • Notebook: /notebooks/translation_module.ipynb
  • Generates teacher-friendly Urdu summaries

5️⃣ Automation & Reporting

  • PDF reports via ReportLab
  • Graphs via Matplotlib & Seaborn (Class Distribution, Confusion Matrix, Accuracy Trend)
  • Auto commit & push to GitHub
  • GitHub Actions for scheduled retraining & report updates

⚙️ System Architecture

flowchart TD
    A[Dataset from Kaggle] --> B[Preprocessing]
    B --> C[ML Model Training]
    C --> D[Evaluation & Metrics]
    D --> E[Visualization + Report Generation]
    E --> F[PDF Report Creation]
    F --> G[Automatic Push to GitHub]
    G --> H[GitHub Actions Retraining Workflow]
Loading

📂 Implementation Summary

Component Description Output
data/ Raw + processed student dataset student_performance.csv
notebooks/ Core model notebooks 5 Notebooks (ML, Forecast, NLP, Translation)
models/ Serialized trained models .pkl files
reports/ Auto-generated PDF & graphs /reports/final_report.pdf
results/ CSVs with evaluation metrics Forecasting & Accuracy results
screenshots/ Saved output images PNGs of runs
README.md Project overview Setup & description
project_report.md Full report This document

📈 Results & Analysis

Classification Performance

Metric Logistic Regression Random Forest Naive Bayes
Accuracy 0.86 0.92 0.84
Precision 0.85 0.93 0.82
Recall 0.84 0.91 0.80
F1-Score 0.84 0.92 0.81

✅ Random Forest achieved the best performance.

Forecasting Results

Metric Value
Mean Absolute Error (MAE) 3.82
R² Score 0.88
Predicted Score Range 50–98

Forecast model predicts future trends accurately.

NLP Translation Example

Input (English) Output (Urdu)
“Student needs improvement in math.” "طالب علم کو ریاضی میں بہتری کی ضرورت ہے۔"

📊 Visualization Outputs

Chart Description
Class Distribution Number of students per performance level
Confusion Matrix Model strengths and weaknesses
Forecast Graph Predicted vs actual performance trends

🤖 Automation Workflow (CI/CD)

  • GitHub Actions triggered on push or scheduled workflow
  • Installs dependencies
  • Downloads dataset via Kaggle API
  • Retrains models automatically
  • Generates updated graphs & PDF reports
  • Commits & pushes back to GitHub

🌍 Social & Educational Impact

Impact Area Description
Accessibility Learning support in Urdu & regional languages
Data Empowerment Helps educators track student performance
Scalability Open-source, extendable for larger platforms
Inclusivity Supports diverse language and cultural backgrounds

⚠️ Limitations

  • Dataset scope limited (no attendance, socioeconomic, behavioral data)
  • Translation accuracy depends on API/network
  • Forecasting assumes stable learning trends

🚀 Future Work

  • Integrate with real-time LMS systems
  • Use deep learning (LSTM, BERT) for prediction improvements
  • Add speech-to-text Urdu learning module
  • Enhance dashboard for educators
  • Deploy as a Progressive Web App (PWA)

📜 References

🧑‍💻 Authors & Credits

  • Main Developer: Mushahid Ali Kazmi
  • Technical Support: Muhammad Abdullah
  • Affiliation: Independent AI Research Initiative, Pakistan

🏁 Conclusion

TaleemAI demonstrates how AI and automation can transform education by:

  • Empowering students with personalized analytics
  • Supporting teachers with actionable insights
  • Bridging language gaps via Urdu/regional translations
  • Maintaining ethical, open-source, and plagiarism-free development

“Education is not just learning facts, but training the mind to think — TaleemAI trains both minds and machines.”