🚖 Smart Dynamic Pricing System

A hybrid Machine Learning + Reinforcement Learning system for dynamic ride pricing

Combining historical price estimation with Q-learning for adaptive pricing strategies

🚀 Quick Start • 📖 Documentation • 🎯 Features • 📊 Demo

---

🎯 Overview

This project implements an intelligent dynamic pricing system for ride-sharing services that combines:

• 📈 Machine Learning: Historical price estimation using median-based algorithms
• 🤖 Reinforcement Learning: Q-learning agent for optimal price multiplier selection
• 🎨 Interactive Interface: Beautiful Streamlit web application for real-time predictions

The system adaptively adjusts ride prices based on demand, supply, vehicle type, and customer loyalty to maximize revenue while maintaining competitive pricing.

✨ Features

🧠 Intelligent Base Price Estimation

• Hierarchical Fallback Strategy: Duration window → Duration bins → Vehicle type → Global median
• Robust Handling: Works with sparse data and missing historical records
• Multi-tier Support: Economy, Premium, and Luxury vehicle pricing

🎯 Advanced Feature Engineering

• Smart Bucketing: Demand (Low/Medium/High/Very High) and Supply (Low/Medium/High) levels
• Loyalty Mapping: Vehicle type → Customer tier (Regular/Silver/Gold)
• State Representation: Optimized for reinforcement learning

🤖 Q-Learning Agent

• Dynamic Environment: Simulates real ride scenarios as episodes
• Price Multipliers: [0.8×, 1.0×, 1.2×, 1.4×] for flexible pricing
• Reward Optimization: Revenue maximization with cost consideration
• Epsilon-Greedy: Balanced exploration vs exploitation

🎨 Interactive Web Application

• Real-time Predictions: Instant price recommendations
• Auto-estimation: Optional historical price calculation
• Visual Feedback: Clear metrics and state information
• Responsive Design: Works on desktop and mobile

🚀 Live Demo

👉 Try the app here: Smart Dynamic Pricing App

🚀 Quick Start

Prerequisites

• Python 3.8 or higher
• pip package manager

Installation

1. Clone the repository

   git clone https://github.com/DivyanshGupta-16/Smart_Dynamic_Pricing.git
   cd dynamic-pricing-strategy

2. Create virtual environment

   # Windows
   python -m venv venv
   venv\Scripts\activate
   
   # macOS/Linux
   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

4. Run the application

   streamlit run app.py

5. Open your browser to http://localhost:8501

📂 Project Structure

dynamic-pricing-strategy/
├── 📁 data/
│   └── dynamic_pricing.csv          # Historical ride data
├── 📁 src/
│   ├── base_price.py               # Base price estimation logic
│   ├── preprocessing.py            # Data preprocessing & state building
│   ├── rl.py                       # RL environment & Q-learning agent
│   └── utils.py                    # Feature engineering utilities
├── 📁 assets/
│   ├── base_price_1x.png          # Demo screenshots
│   ├── base_price_1-2x.png
│   ├── base_price_1-4x.png
│   └── estimated_base_price.png
├── app.py                          # Streamlit web application
├── requirements.txt                # Python dependencies
└── README.md                       # This file

📊 Demo

🖼️ Application Screenshots

Base Price Estimation	High Demand Scenario
Supply-Demand Imbalance	Duration-based Pricing

📈 Example Output

Base Price (ML Estimate): ₹120.00
RL Multiplier: 1.20×
State: ('High', 'Medium', 'Silver')
─────────────────────────────────
💰 Final Recommended Price: ₹144.00

🧠 Methodology

🔄 Hybrid Approach

1. Machine Learning Component

   • Analyzes historical ride data
   • Estimates base price using statistical methods
   • Handles missing data and edge cases

2. Reinforcement Learning Component

   • Q-learning agent learns optimal pricing strategies
   • Considers demand, supply, and customer loyalty
   • Maximizes revenue through dynamic multiplier selection

3. Integration

   • Final Price = Base Price × RL Multiplier
   • Real-time adaptation to market conditions

📊 Data Requirements

Your dataset should include these columns:

Number_of_Riders,Number_of_Drivers,Vehicle_Type,Expected_Ride_Duration,Historical_Cost_of_Ride

🛠️ Usage

Web Interface

1. Adjust sliders for demand (riders) and supply (drivers)
2. Select vehicle type and expected duration
3. Choose between manual input or auto-estimation for base price
4. View real-time price recommendations

Programmatic Usage

from src.base_price import estimate_base_price
from src.rl import train_q_agent, DynamicPricingEnv
from src.preprocessing import load_data, preprocess

# Load and preprocess data
df = load_data("data/dynamic_pricing.csv")
processed_df = preprocess(df)

# Train RL agent
env = DynamicPricingEnv(processed_df)
agent = train_q_agent(env, episodes=25)

# Get price recommendation
base_price = estimate_base_price(df, "Premium", 30)
multiplier = agent.get_best_action(state)
final_price = base_price * multiplier

📋 Dependencies

• streamlit ≥ 1.33 - Web application framework
• pandas ≥ 2.0 - Data manipulation and analysis
• numpy ≥ 1.24 - Numerical computing

---

⭐ Star this repository if you found it helpful!

🔝 Back to Top