Product Demand Forecast App

A professional, machine learning-powered web application designed to forecast future product demand based on historical sales data. Built with Python and Streamlit, the application features a clean, high-performance developer dashboard and utilizes a Random Forest Regressor to provide accurate time-series demand predictions.

Key Features

• Executive Dashboard: Unified overview featuring 5 KPI metrics (Dataset status, Total records, Forecast period, Model type, and Status), interactive charts, and highlights.
• Dynamic Data Ingestion: Custom CSV file upload interface alongside a fallback feature to load a default demo dataset instantly.
• Interactive Forecasting: Category-specific prediction tools with an adjustable monthly horizon slider (1 to 12 months) and tabular forecast outputs.
• Sales Analytics: Deep dive insights including total sales breakdowns and ranking of the top 10 product categories.
• Tabular Data Preview: Interactive data explorer to filter, view, and inspect raw historical records.
• Export Capabilities: Generate printable summaries and export forecasted demand lists directly as CSV files.
• Premium Interface: A clean, modern light-themed UI utilizing custom CSS layouts, sidebar navigation, and material vector icons.

Tech Stack

• Frontend Framework: Streamlit
• Machine Learning: Scikit-Learn (Random Forest Regressor)
• Data Manipulation: Pandas, NumPy
• Data Visualization: Matplotlib
• Typography & Styling: Google Fonts (Outfit), Custom CSS

Installation and Setup

Prerequisites

Ensure you have Python 3.9+ installed on your system.

1. Clone the Repository

git clone https://github.com/vasanth642/Product-Demand-Forcast.git
cd Product-Demand-Forcast

2. Install Dependencies

Install the required Python libraries using pip:

pip install -r requirements.txt

3. Run the Application

Start the Streamlit server locally:

streamlit run app.py

The application will open automatically in your browser at http://localhost:8501.

File Structure

Product-Demand-Forcast/
├── app.py                     # Main Streamlit application and layout
├── requirements.txt           # Python package dependencies
├── monthly_store_sales.csv    # Demo sales dataset
└── README.md                  # Project documentation

System Architecture & Data Flow

The diagram below details the end-to-end data pipeline from importing raw historical sales data to generating and downloading predicted demand values.

graph TD
    A[User Uploads CSV / Loads Demo Data] --> B[Data Ingestion & Validation]
    B --> C[Data Preprocessing & Feature Engineering]
    C -->|Extracts Month, Year, Time Index| D[Train Random Forest Regressor Model]
    D --> E[Generate Future Timestamps]
    E --> F[Inference: Predict Future Demand]
    F --> G[Dashboard Visualization & Analytics]
    F --> H[Download Predictions as CSV]

Loading

Stage-by-Stage Description

1. Data Ingestion & Validation

   • The user imports sales records through the file uploader or uses the pre-configured fallback demo dataset (monthly_store_sales.csv).
   • The system validates that the dataset contains the required schema: Month (Date), family (String/Product Type), and sales (Numeric/Volume).

2. Preprocessing & Feature Engineering

   • The system parses raw date strings into Datetime objects and chronologically sorts the records.
   • For the selected product category, the system generates time-series features:
     • month_num (Extracts calendar month 1-12 to capture seasonality).
     • year (Extracts calendar year to track long-term trend progression).
     • time_index (A sequential index representing elapsed time periods).

3. Model Training

   • A Scikit-Learn RandomForestRegressor (with 200 estimators) is trained using the engineered features (time_index, month_num, year) as inputs to fit target sales data.
   • The training step utilizes Streamlit's resource-caching system (@st.cache_resource) to prevent redundant model fitting when switching views.

4. Inference & Prediction

   • The system builds a future feature matrix based on the selected forecasting horizon (e.g. 6 months).
   • The trained model performs inference to predict future monthly demand numbers.

5. Visualization & Export

   • Visualizations are rendered as clean, high-contrast trend charts using styled Matplotlib figures.
   • Users can filter predictions, review metrics, preview underlying source metadata, and download predictions as a standardized CSV report.