M&A Prediction Pipeline

Objective: Predict M&A target events using machine learning on fundamental financial data.

📂 Directory Structure

MnA_Prediction/
├── README.md
├── src/                           # Source code
│   ├── mna_colab_pipeline.py      # Main Colab notebook
│   └── feature_engineering.py     # Original reference script
│
├── data/                          # Data files (on GitHub)
│   ├── deals/
│   │   ├── dma_corpus_metadata_with_factset_id.csv  # 2000-2020
│   │   └── factset_xls/                             # 2000-2025
│   │       ├── 2000to05Batch1.xls
│   │       └── ...
│   └── fundamentals/
│       └── compustat_funda_2000on.csv
│
└── archive/                       # Old/unused files

🚀 Quick Start (Google Colab)

1. Open src/mna_colab_pipeline.py in Google Colab
2. Ensure your Google Drive contains:
   • fundq_full.parquet (quarterly Compustat, ~564 MB)
   • funda_full.parquet (annual Compustat, ~200 MB)
3. Run all cells

Data Sources

Source	Location	Coverage
Compustat (fundq/funda)	Google Drive	Through ~2020
DMA Corpus	data/deals/	2000-2020
FactSet XLS	data/deals/factset_xls/factset_2000_2025/	2000-2025

Pipeline Features

• Multi-horizon labeling (3m-24m targets)
• Probability calibration (prior correction + isotonic)
• Event study verification
• S&P 500 benchmarking
• Schema-preserving data extension

License

MIT