DataFusion - Python Application

DataFusion

📊 A CSV workbench that runs on the desktop — load, clean, encode, scale and plot a dataset without writing a line of pandas

DataFusion is a desktop application for the part of data science that comes before the model: opening a file, finding out what is wrong with it, and fixing it. Load a CSV or one of the two bundled UCI datasets, read the statistics, fill or drop the missing values, encode the categorical columns, scale the numeric ones, and draw the chart that shows whether any of it worked.

Every transformation happens on a working copy, and one button puts the original data back — so exploring destructively is safe.

🎯 Key Features

• Undo is a first-class button — the original frame is kept beside the working one, so any cleaning step can be reversed without reloading the file.
• Descriptive statistics on demand — count, mean, spread, plus skewness and kurtosis from SciPy, which are the two that actually tell you whether a distribution will behave.
• Both correlation matrices, side by side — Pearson and Spearman together, because a monotonic relationship that is not linear is exactly what one of them hides.
• Missing values handled by strategy, not by hand — pick a strategy and it is applied to the frame; duplicates are removed the same way.
• Three encodings for categorical columns — one-hot, binary and target encoding, chosen per column rather than applied blindly to everything.
• Scaling on selected columns — standard or min-max, applied to the columns you choose, so an identifier column is not scaled along with the measurements.
• Five chart types, each configurable — histogram, box plot, bar chart, line plot and pie chart, drawn with Matplotlib into the window and saveable as an image.
• Sub-tables by row and column index — extract or exclude a slice and keep working on it.
• Two datasets included — UCI Adult and Chronic Kidney Disease, ready to open without hunting for a file.

🖼️ Screenshots

The main view	Descriptive statistics

Cleaning and transformation	Correlation matrices

Charts	Scaling results

📚 Bundled Datasets

Both live in database/, with their original .names and .info files, so the column meanings are not lost.

🎬 Demo

🛠️ Technology Stack

🚀 Getting Started

Prerequisites

• Python 3.10 or newer
• Tkinter available in the interpreter (it ships with most CPython builds)

1. Clone the repository

git clone https://github.com/dawidolko/DataFusion-App-Python.git
cd DataFusion-App-Python

2. Install dependencies

pip install -r src/requirements.txt

3. Run

python src/main.py

The window opens with a start screen; Go to data loads either a bundled dataset or a CSV of your own.

📁 Project Structure

DataFusion-App-Python/
├── src/
│   ├── main.py             # the whole application: window, tabs, transformations, charts
│   └── assets/             # icons and the screenshots used above
├── database/
│   ├── adult/              # UCI Adult: data, test split, column names
│   └── chronic/            # Chronic Kidney Disease in ARFF, with its info file
└── docs/                   # documentation, task descriptions

♿ Accessibility

An accessibility panel in the corner of every page changes text size, contrast, font and motion, and remembers the choice on the device. It sits on top of the markup and does not replace accessible markup.

📄 License

The application code is MIT © Dawid Olko. Dependencies keep their own licences, listed in CREDITS.md.