Bank Statement to QIF Converter

Convert bank statements from PDF to QIF format using supervised learning from existing PDF-QIF pairs.

Features

• PDF to QIF Conversion
  • Process bank statements in PDF format
  • Support for multiple bank accounts
  • Batch processing of multiple statements
• Supervised Learning Approach
  • Learn from existing PDF-QIF pairs
  • Match transactions based on dates, amounts, and descriptions
  • Support multiple PDF statements mapped to a single QIF file per bank account
• Smart Transaction Matching
  • Weighted scoring system for accurate matching
  • Preserves original transaction details
  • Handles various date and amount formats
• QIF Export
  • Generates standard QIF files
  • Compatible with most financial software
  • Maintains transaction integrity

Setup

1. Create a conda environment:

conda create -n bank_processor python=3.11
conda activate bank_processor

2. Install dependencies:

pip install -r requirements.txt

Usage

Directory Structure

Organize your files as follows:

your_working_directory/
├── statements/          # Your bank statements (PDF)
│   ├── account1/       # Statements for first account
│   │   ├── jan2024.pdf
│   │   ├── feb2024.pdf
│   │   └── ...
│   └── account2/       # Statements for second account
│       ├── jan2024.pdf
│       └── ...
├── training/           # Training QIF files (one per account)
│   ├── account1.qif
│   └── account2.qif
└── output/            # Where new QIF files will be saved

Processing Multiple Files

Create an input list file (e.g., input_files.txt) with the following format:

# Training files (PDF and QIF pair)
statements/account1/dec2023.pdf
training/account1.qif

# Files to process
statements/account1/jan2024.pdf
statements/account1/feb2024.pdf
statements/account1/mar2024.pdf
output/account1_q1_2024.qif

The file should contain:

1. Training section with:
   • Path to the training PDF statement
   • Path to the corresponding QIF file
2. Processing section with:
   • Paths to all PDF files to process
   • Path to the output QIF file (last line)

Then run:

python src/main.py --input-list input_files.txt

Processing Single Files

For individual files:

python src/main.py --train-pdf TRAINING_PDF --train-qif TRAINING_QIF --pdf NEW_PDF --output OUTPUT_QIF

Example:

python src/main.py --train-pdf statements/account1/dec2023.pdf --train-qif training/account1.qif --pdf statements/account1/jan2024.pdf --output output/jan2024.qif

Transaction Matching

The system uses a sophisticated matching algorithm that considers:

• Transaction dates (weight: 3)
• Transaction amounts (weight: 3)
• Transaction descriptions (weight: 2)

A match is considered valid when the similarity score exceeds 0.7 (70% similarity).

Notes

• Each bank account should have its own training QIF file
• Multiple PDF statements can be mapped to a single training QIF
• The system preserves original transaction dates and amounts
• Categories and other metadata are copied from the best matching transaction in the training data
• If no match is found for a transaction, it will be included without categorization
• The system supports various date formats and handles different amount representations
• All processing is done locally - no data is sent to external services

Error Handling

• Invalid PDFs or unreadable files will be reported with appropriate error messages
• Transactions without matches will be included in the output but flagged
• Processing continues even if some transactions can't be matched
• Detailed logs help identify any issues during processing