Skip to content

README.md

Latest commit

 

History

History
125 lines (95 loc) · 4.28 KB

README.md

File metadata and controls

125 lines (95 loc) · 4.28 KB

Exam Analytics Pro: Agentic Difficulty Prediction System 🎓🤖

Streamlit App Python 3.12+ LangGraph

Exam Analytics Pro is a high-performance system designed to predict and analyze the difficulty of assessment items using a hybrid approach of Traditional Machine Learning and Agentic AI (LangGraph/LangChain).

🌟 Key Features

1. Hybrid Intelligence Engine

  • Vectorized Inference: Uses TF-IDF and specialized scikit-learn models (Logistic Regression & Decision Trees) for fast, data-driven baseline predictions.
  • Agentic Reasoning: Employs a LangGraph state machine to orchestrate multi-step analysis by LLMs (OpenAI, Gemini, Groq).
  • Pedagogical Analysis: Provides deep insights into Bloom's Taxonomy levels, cognitive depth, and linguistic complexity.

2. Multi-Provider LLM Support

Switch between state-of-the-art LLM providers seamlessly:

  • Google Gemini: Reliable and fast pedagogical analysis.
  • Groq (Llama 3): Ultra-low latency inference for real-time reasoning.
  • OpenAI (GPT-4o): High-fidelity conceptual complexity mapping.

3. Interactive Analytics Dashboard

  • CSV Batch Processing: Upload hundreds of questions for instant difficulty distribution reports.
  • Advanced Insights: Correlate question length, tag count, and engagement scores with predicted difficulty.
  • Explainable AI: Click a button to see exactly why an AI agent classified a question a certain way.

🏗️ System Architecture

graph TD
    A[Raw Exam Dataset] -->|CSV Upload| B[Dashboard]
    B -->|Preprocessing| C[Feature Matrix]
    
    subgraph Agentic_Workflow
        direction TB
        Start((START)) --> ML_Predictor[ML classification Node]
        ML_Predictor -->|Baseline: Hard/Med/Easy| LLM_Reasoner[LLM Reasoning Node]
        LLM_Reasoner -->|Contextual Analysis| Synthesizer[Synthesis Node]
        Synthesizer --> End((END))
    end
    
    C --> Start
    ML_Predictor -.-> Models[(scikit-learn Models)]
    LLM_Reasoner -.-> APIs{LLM APIs}
    
    End -->|Final Verdict| B
    B -->|UI Output| Visualization[Distribution Charts & Reports]
Loading

🚀 Getting Started

Prerequisites

  • Python 3.12+
  • Virtual Environment (recommended)

Installation

  1. Clone the repository:

    git clone https://github.com/praggCode/Exam-Analytics-System.git
cd Exam-Analytics-System

  2. Set up virtual environment:

    python3 -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

  3. Install dependencies:

    pip install -r requirements.txt

Running the Application

  1. Start the Streamlit Server:

    streamlit run ui/app.py

  2. Configure AI Settings:

    • Open the app at http://localhost:8501.
    • In the sidebar, select your preferred LLM Provider (Groq, Gemini, or OpenAI).
    • Enter your API Key.

  3. Analyze Data:

    • Upload a CSV containing a question column.
    • Click "Analyze Difficulty" to see ML results.
    • Use the "Generate AI Reasoning" button for a deep-dive analysis of your top questions.

📁 Project Structure

├── data/               # Processed and raw datasets
├── src/
│   ├── ai_engine.py    # LangGraph workflow definition
│   ├── model_train.py  # Model training pipeline
├── models/             # Saved joblib models (.joblib)
├── ui/
│   ├── app.py          # Streamlit UI implementation
├── requirements.txt    # Project dependencies
└── README.md           # Project documentation

🛠️ Technology Stack

  • Dashboard: Streamlit
  • ML Framework: scikit-learn, joblib
  • Orchestration: LangChain, LangGraph
  • NLP: Beautiful Soup (BS4), re, TF-IDF
  • Data Science: Pandas, NumPy, Seaborn, Matplotlib

📝 License

This project is for educational and professional assessment analytics. Built with ❤️ and AI.

For a detailed report on the LangChain implementation, see langchain_report.md.