Frankenstein

A small Python toolkit for loading, cleaning, analyzing, clustering, and visualizing experimental datasets.

Key Features

• Load Excel (.xlsx) files into a lightweight Dataset API
• Immutable DataInstance/Dataset abstractions with schema integrity handling
• Cleaning utilities to normalize values/keys (e.g., unify NaNs, fix typos)
• Plotting helpers:
  • Correlation scatter with stats and best-fit line
  • Category comparison (box + strip) with stats annotation
• Clustering grid search across preprocessing, dimensionality reduction (t-SNE/UMAP), and clustering methods (KMeans, Spectral, GMM)
• Saves figures to plots/ and a grid-search summary Excel

Tech Stack

• Python 3.11+
• numpy, pandas
• matplotlib, seaborn
• scipy
• scikit-learn
• umap-learn
• openpyxl (for reading .xlsx via pandas)

Project Structure

.
├─ analysis_demo.py           # Example workflow: load → clean → analyze → plot → cluster
├─ data/                      # Local data (Excel files). Consider ignoring in Git
│  └─ DataFromJoe/            # Example source datasets (.xlsx)
├─ plots/                     # Auto-generated figures and summaries
└─ src/
   ├─ analysis/
   │  └─ clustering.py        # Preprocess, reduce dimensions, and cluster (+ grid search)
   ├─ plotter/
   │  ├─ basic_plotting.py    # Correlation/comparison plots with stats overlays
   │  ├─ clustering_plotting.py
   │  └─ stats_plotting.py
   ├─ data_loader.py          # Excel loader → Dataset
   ├─ data_cleaner.py         # Cleaning rules (e.g., joe_data_cleaner)
   └─ data_core.py            # DataInstance/Dataset abstractions

Usage

Run the demo

python analysis_demo.py

This will:

• Load multiple Excel files into a single Dataset
• Apply cleaning rules (joe_data_cleaner)
• Produce correlation and comparison plots
• Run a clustering grid search (preprocessing × dimension reduction × clustering)
• Save figures under plots/ and a grid-search summary Excel at plots/clustering/grid_search_summary.xlsx