🔮 Demand Forecasting App

A powerful Streamlit-based web application for demand forecasting that combines the strengths of Prophet and N-Beats models, enhanced with Gemini AI for intelligent pattern detection and analysis.

✨ Features

• 🔄 Multi-Model Forecasting:

  • Prophet: Meta's time series forecasting tool that handles seasonality and holiday effects
  • N-Beats: Deep neural architecture specifically designed for time series forecasting with interpretable outputs

• 📊 Interactive Analysis:

  • Upload CSV files with historical sales data
  • Visualize forecasts with interactive charts
  • Compare different forecasting models

• 🤖 AI-Powered Insights:

  • Gemini Bot for image-based trend analysis (forecast visualizations)
  • Gemini Bot for data-based Q&A (CSV analysis)

• 💾 Export Options:

  • Download forecasts as CSV
  • Save forecast visualizations as images

🧠 Understanding the Models

Prophet

Prophet is a forecasting procedure developed by Meta (formerly Facebook) that:

• Automatically detects seasonal patterns (daily, weekly, yearly)
• Handles missing data and outliers gracefully
• Accommodates holiday effects and special events
• Uses a decomposable model with trend, seasonality, and holiday components
• Is particularly effective for business forecasting with multiple seasonalities

N-Beats (Neural Basis Expansion Analysis for Time Series)

N-Beats is a deep learning approach that:

• Uses deep neural networks specifically designed for time series
• Provides interpretable forecasts through trend and seasonality decomposition
• Often outperforms traditional statistical methods for complex patterns
• Can capture non-linear relationships in your data
• Works well even with limited historical data

🛠️ Tech Stack

• Backend: Python
• ML Models: Prophet, N-Beats (TensorFlow/Keras)
• AI Integration: Google Gemini API
• Frontend: Streamlit
• Data Processing: Pandas, NumPy
• Visualization: Matplotlib

🎬 Demo

https://github.com/user-attachments/assets/9fc5bf11-8f60-4780-8e4d-2280cb09943a

⚙️ Installation

1. Clone the repository:

   git clone https://github.com/yourusername/demand-forecasting-app.git
   cd demand-forecasting-app

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

4. Set up environment variables:

   • Create a .env file in the root directory
   • Add your Google API key:

     GOOGLE_API_KEY=your_api_key_here
     

5. Run the application:

   streamlit run multipage.py

🚀 Usage

Generate Forecasts

1. Upload a CSV file with Date and Sales columns
2. Select forecast end date
3. Choose between Prophet or N-Beats model
4. View and download forecasts

Chat with Image

1. Upload forecast visualization images
2. Get AI-powered analysis of trends and patterns

Chat with Data

1. Upload your sales data (CSV or Excel)
2. Ask questions about trends, anomalies, and insights

📋 Data Requirements

CSV files must contain:

• A Date column (formats supported: DD-MM-YYYY, MM/DD/YYYY, etc.)
• A Sales column with numerical values

Date,Sales
01/01/2023,100
01/02/2023,150
01/03/2023,200

📂 File Structure

demand-forecasting-app/
├── main.py               # Main forecasting functionality
├── multipage.py          # Multi-page app configuration
├── image_bot.py          # Image analysis with Gemini
├── data_bot.py           # Data analysis with Gemini
├── prophet_script.py     # Prophet model utilities
├── nbeats.py             # N-Beats model implementation
├── .env.example          # Environment variables template
├── requirements.txt      # Dependencies
└── README.md             # This file

📸 Screenshots

Project Map

Forecast Generation

Graph Analysis

Data Analysis

🤔 When to Use Each Model

👤 Author

For any questions or issues, please open an issue on GitHub: @Siddharth Mishra

Made with ❤️ and lots of ☕