sahayabhineet/Code-Complexity-Predictor

Code Complexity Predictor

A machine learning web application that predicts the time complexity of code snippets using Natural Language Processing and Logistic Regression.

Features

• Real-time Prediction: Analyze code complexity instantly
• Multiple Complexity Classes: Predicts Constant, Linear, Quadratic, and Cubic complexities
• Confidence Scores: Shows probability distribution across all complexity classes
• Simple UI: Clean, intuitive interface for easy code analysis
• RESTful API: JSON API endpoint for integration with other tools

Technologies Used

• Backend: Flask (Python)
• Machine Learning: scikit-learn, TF-IDF Vectorization, Logistic Regression
• Dataset: CodeParrot CodeComplex dataset
• Frontend: HTML, CSS, JavaScript
• Deployment: Gunicorn, Docker

Usage

Web Interface

1. Paste your code into the text area
2. Click "Analyze Complexity"
3. View the predicted complexity and confidence scores

API Endpoint

POST /predict

Request Body:

{
  "code": "your code here"
}

Response:

{
  "complexity": "linear",
  "probabilities": {
    "constant": 0.15,
    "linear": 0.65,
    "quadratic": 0.15,
    "cubic": 0.05
  }
}

Model Details

• Algorithm: Logistic Regression
• Feature Extraction: TF-IDF Vectorization (max 5000 features)
• Dataset: CodeParrot CodeComplex dataset
• Train/Test Split: 80/20
• Complexity Classes: Constant, Linear, Quadratic, Cubic