🌿 Prakriti Nishchikaran: Machine Learning for Ayurvedic Body Constitution Analysis

📖 Table of Contents

• Introduction
• Motivation
• Team members
• What is Prakriti?
• Features
• Technical Implementation
• API Usage
• Model Evaluation
• Web Application
• Applications & Future Scope
• Acknowledgements

---

🌟 Introduction

Prakriti Nishchikaran is an innovative machine learning system that analyzes 40+ bio-behavioral traits to determine an individual's Ayurvedic body constitution (Prakriti) with 88.5% accuracy. This project bridges ancient Ayurvedic wisdom with modern technology, providing a standardized, scalable solution for Prakriti analysis.

🔮 Accuracy: 88.5%
🚀 Live: link

---

👨‍🏫 Team Members

Name	Role	GitHub Username
Dr. Krantee Jamdaade	Mentor, Dataset Contributor	Not on GitHub
Benjamin Sebastian	Full Stack Developer, ML Model	@SebastianBenjamin
Arunima Pandey	Research & Frontend Developer	@Arunima04

---

💡 Motivation

The project was inspired by the story of Rajesh, a software engineer who suffered health collapse due to work stress. Traditional Ayurvedic diagnosis faced several challenges:

• Practitioner-dependent results with high variability
• Limited accessibility requiring in-person consultations
• Lack of standardization in assessments
• Scarcity of experts

Our solution addresses these by:

• Providing objective analysis through machine learning
• Creating a digital platform accessible anywhere
• Developing quantifiable metrics for human traits
• Delivering instant results with high accuracy

---

🧘 What is Prakriti?

Prakriti refers to an individual's inherent mind-body constitution, determined by the unique combination of three doshas:

• Vata: Associated with movement and creativity
• Pitta: Governs metabolism and transformation
• Kapha: Provides structure and lubrication

Each person has a distinct ratio of these doshas that remains constant throughout life, serving as a unique biological blueprint for health and wellness.

---

✨ Features

• Advanced ML Model: XGBoost classifier trained on 55+ bio-behavioral traits
• Standardized Framework: Quantitative analysis of traditionally qualitative parameters
• Cloud Hosting:
  • API hosted on Render
  • Frontend interface on Netlify
• Comprehensive Output:
  • Primary Prakriti prediction
  • Probabilities for all dosha combinations
  • Integration with Google Gemini for detailed interpretations

---

💻 Technical Implementation

🛠️ Model Development Journey

1. Data Collection:

   • Initial dataset with 71 columns from Ayurvedic research
   • Added encoded Prakriti column (1=Pittaj, 2=Vataj, 3=Kaphaj)

2. Feature Selection:

   • Started with 5 key features:
     • Usual body temperature
     • Appetite
     • Sleep patterns
     • Emotion
     • Physique
   • Expanded to 55+ features after initial testing

3. Model Comparison:

   • Tested multiple algorithms:
     • Random Forest
     • KNN
     • Decision Tree
     • XGBoost
     • Voting Classifier combinations
   • XGBoost selected for its sequential learning and feature dependency handling

4. Preprocessing:

   • SMOTE for handling class imbalance
   • Label Encoding for target variable
   • Standard Scaling for feature normalization

📊 Model Architecture

graph TD
    A[Raw Data] --> B[Feature Selection]
    B --> C[Preprocessing]
    C --> D[Model Training]
    D --> E[XGBoost Classifier]
    E --> F[API Deployment]

Loading

---

📡 API Usage

🔗 Endpoints

GET / (Root URL)
https://prakruthinischikaran-api.onrender.com

POST /predict
https://prakruthinischikaran-api.onrender.com/predict

📥 Sample Input

{
  "Skin color": 1,
  "Skin texture": 1,
  "Hair quality": 1,
  ... [40+ additional features]
}

📤 Sample Response

{
  "prediction": "Vataj",
  "probabilities": {
    "Kaphaj": 0.0068,
    "Kaphaj-Pittaj": 0.0069,
    "Pittaj": 0.0458,
    "Pittaj-Kaphaj": 0.0818,
    "Pittaj-Vataj": 0.0072,
    "Vataj": 0.7019,
    "Vataj-Pittaj": 0.1493
  }
}

🧪 Test Script

import requests

url = "https://prakruthinischikaran-api.onrender.com/predict"
data = {
    "Skin color": 1,
    "Skin texture": 1,
    # ... [complete feature set]
}

response = requests.post(url, json=data)
print("Status Code:", response.status_code)
print("Response:", response.json())

---

📈 Model Evaluation

🧠 Feature Importance

Top influential features in the XGBoost model

📊 Classification Metrics

Precision, recall, and f1-scores for each dosha class

🔄 Confusion Matrix

Model performance across different Prakriti types

---

🌐 Web Application

The system includes a user-friendly web interface with:

• Interactive form for trait input
• Visual results dashboard showing dosha distribution
• Gemini AI integration for personalized recommendations
• Responsive design accessible on all devices

---

🚀 Applications & Future Scope

Current Applications:

• Clinical Use: Ayurvedic practitioners can use for preliminary analysis
• Personal Wellness: Individuals can understand their constitution
• Integrated Therapy: Basis for personalized aromatherapy recommendations

Future Enhancements:

1. Hardware Integration: Already implemented prototype of Ayur Aroma Diffuser
2. Multimodal Recommendations:
   • Custom aromas based on Prakriti
   • Visual therapy suggestions
   • Raga (musical) therapy prescriptions
3. Health Monitoring: Longitudinal tracking of dosha balance
4. Mental Health: Prakriti-based stress and anxiety management

---

🙏 Acknowledgements

Developed under the guidance of Dr. Krantee Jamdaade at KJSIM Data Science Department. Special thanks to our research team for bridging Ayurvedic wisdom with machine learning technology.

"Where ancient wisdom meets modern science for holistic well-being"