---

Automated Exploratory Data Analysis Web Application

---

📌 Project Overview

AUTO-EDA is an automated exploratory data analysis tool built to extract meaningful insights from any CSV dataset — instantly.

The objective of this project is to replicate the core functionality of ydata-profiling through a clean, interactive Streamlit web interface built from scratch using Python.

The complete workflow covers automatic column type detection, statistical profiling, correlation analysis, missing value detection, and dynamic visual reporting — all without writing a single line of analysis code.

---

📈 Application Preview

Dashboard Interface

Overview

Variables

Interaction

Correlation

Missing Values

Sample

---

🎯 Problem Statement

Raw CSV datasets require significant manual effort to understand — column types, distributions, correlations, and missing values all need separate investigation.

The challenge was to:

• Automatically detect column types (Numeric, Categorical, Text)
• Extract meaningful statistical metrics per column
• Identify correlations across all variable types
• Detect and visualize missing value patterns
• Build an interactive, multi-page analytical interface

This project focuses on transforming raw tabular data into structured analytical insights through automated pipelines.

---

📊 Dataset

• Format: .csv (Any user-uploaded CSV file)
• Type: Structured tabular data
• Source: User-uploaded via Streamlit sidebar
• Supports: Numeric, Categorical, and Text columns

Note: No dataset is stored in this repository. Upload your own CSV to explore.

---

⚙️ Tools & Technologies Used

• Python – Core programming language
• Pandas – Data cleaning, transformation, aggregation
• Matplotlib – Statistical plotting and chart generation
• Seaborn – Heatmaps and advanced visualizations
• WordCloud – Text frequency visualization
• Missingno – Missing value matrix and heatmap
• Scikit-learn – Label encoding for correlation analysis
• Streamlit – Interactive multi-page web application

---

🧱 Workflow Architecture

Raw CSV Upload
→ Column Type Detection (Numeric / Categorical / Text)
→ Statistical Profiling (per column)
→ Alert Generation (Missing, Zeros, Unique, Correlation)
→ Interaction & Correlation Analysis
→ Missing Value Visualization
→ Interactive Dashboard (Streamlit)

---

📊 Analytical Features Implemented

• Overview — Dataset statistics, variable type breakdown, smart alerts, reproduction info
• Variables — Per-column profiling with stats table and type-appropriate chart
• Interactions — Scatter plot between any two numeric columns
• Correlations — Heatmap and correlation table (encodes categorical columns automatically)
• Missing Values — Count bar chart, nullity matrix, and nullity heatmap
• Sample — Head, tail, and random sample of the dataset

---

🚨 Smart Alerts

AUTO-EDA automatically flags:

• ⚠️ Columns with missing values (count + percentage)
• 🔵 Numeric columns with high zero count
• 🔴 Columns with all unique values
• 🔴 Uniformly distributed numeric columns
• ⚫ Highly correlated column pairs (threshold > 0.5)

---

📂 Project Structure

AUTO-EDA/
│
├── app.py                  → Main Streamlit Application
│
├── utils/
│   ├── overview.py         → Dataset statistics & column classification
│   ├── variables.py        → Column type detection logic
│   ├── plots.py            → All plot/chart generation functions
│   ├── correlation.py      → Correlation heatmap logic
│   └── missing.py          → Missing value analysis functions
│
├── requirements.txt        → Dependencies
└── images/                 → Screenshots and architecture diagrams

---

▶ How to Run the Application

1. Clone the repository:

git clone https://github.com/<your-username>/auto-eda.git

2. Install dependencies:

pip install -r requirements.txt

3. Run the Streamlit app:

streamlit run app.py

4. Upload any CSV file from the sidebar and explore your data instantly.

---

📦 Requirements

streamlit
pandas
matplotlib
seaborn
wordcloud
missingno
scikit-learn

Install all at once:

pip install -r requirements.txt

---

💡 Key Learning Outcomes

• Automatic column type classification using statistical heuristics
• Replicating ydata-profiling functionality from scratch
• Encoding categorical and text variables for correlation analysis
• Building multi-page Streamlit applications with modular utility functions
• Type-aware visualization selection (histogram / bar chart / wordcloud)
• Missing value analysis using missingno library

---

🔗 Important Links

🚀 Live Streamlit Application

Access the deployed app here: Click Here to access

📖 Medium Blog (Detailed Project Explanation)

Read the full case study here: Click Here to read

📊 Project Presentation (PPT Slides)

View the presentation slides here: [Add Presentation Link Here]

🎥 YouTube Walkthrough (Application Demo & Explanation)

Watch the full project demo here: [Add YouTube Demo Link Here]

---

Made with ❤️ by Pratham Harer • AUTO-EDA