WinsorWeb – Flask-based Winsorization & outlier detection app

📊 Overview

A Flask-based data science application for statistical outlier detection, visualization, and cleaning.
The app provides interactive insights into normal distribution analysis using the Empirical Rule (68-95-99.7) and applies Winsorization for robust outlier treatment.

---

🚀 Features

• Outlier Detection (Z-Score Method):
  • Identifies extreme values beyond ±3σ using normal distribution properties.
  • Provides statistical insights (Mean, Standard Deviation, Standard Error).
• Interactive PDF Plot:
  • Visualizes probability density function with highlighted outliers.
  • Built using Plotly for dynamic graph rendering.
• Winsorization (Robust Data Cleaning):
  • Applies 5% Winsorization on dataset for treating extreme values.
  • Displays original vs cleaned data side by side.
• Web Application (Flask):
  • Clean UI with separate pages for analysis and cleaned dataset view.

---

🛠️ Technologies & Libraries Used

• Flask → Web framework for routing and UI integration.
• Pandas → Data manipulation and analysis (outlier.csv, dataset.csv).
• NumPy → Numerical computations, percentiles, and array processing.
• SciPy (stats) → Normal distribution (PDF), SEM, and Z-score calculation.
• Plotly Graph Objects → Interactive probability density function plots.
• Plotly I/O → Conversion of plots into HTML for Flask rendering.
• OS Module → File handling and dataset existence checks.

---

📈 Workflow

🔹 Home Page (Outlier Detection & Normal Distribution)

1. Load dataset (outlier.csv).
2. Compute statistics → Mean, Standard Deviation, Standard Error.
3. Apply Empirical Rule (68-95-99.7) to interpret spread.
4. Detect outliers using Z-Score > ±3.
5. Visualize via interactive PDF plot with highlighted outliers.

🔹 Clean Data Page (Winsorization)

1. Load dataset (dataset.csv) or generate dummy dataset if not found.
2. Apply 5% Winsorization → clip values below 5th percentile & above 95th percentile.
3. Display side-by-side comparison: Original vs Winsorized values.

---

📂 Project Structure

├── app.py                # Main Flask application
├── templates/
│   ├── index.html        # Homepage for outlier detection
│   ├── clean_data.html   # Winsorized dataset view
├── static/               # CSS, JS, images if required
├── dataset.csv           # Input dataset for Winsorization (optional)
├── outlier.csv           # Input dataset for outlier detection
└── README.md             # Project documentation

---

⚙️ Installation & Usage

1. Clone the repository:

   git clone https://github.com/Imswappy/WinsorWeb.git
   cd WinsorWeb

2. Run Flask app:

   python app.py

3. Open browser and navigate to:

   http://127.0.0.1:5000/
   

---

📊 Example Outputs

• PDF Plot with Outliers Highlighted

• Table of Outliers

• Winsorized Dataset Table
• 

---

✅ Conclusion

This application combines statistics, visualization, and robust data cleaning into a seamless web interface.
It is particularly useful for data preprocessing, anomaly detection, and exploratory data analysis (EDA) in machine learning pipelines.