Rabbit

Financial Data Processing & Reporting System

Executive Summary

A web-based financial data processing system developed for an accounting practice to reduce manual transaction processing and improve reporting efficiency. The system extracts transaction data from bank statements, converts unstructured financial records into a structured database, and provides reporting and filtering capabilities that support accounting and financial analysis workflows. The solution was designed to address the time-consuming and error-prone process of manually capturing transaction data from bank statements.

Business Problem

Business Challenge

Accountants received bank statements in PDF format.

Staff had to:

1. Open statements
2. Read transactions manually
3. Capture information into spreadsheets
4. Categorize transactions
5. Prepare reports

Challenges:

• Time-consuming
• Repetitive
• Human error risk
• Difficult to scale

Requirements Analysis

Stakeholder Requirements

The accounting firm required:

• Secure document uploads
• Automated transaction extraction
• Transaction categorization
• Search and filtering capabilities
• Reporting functionality
• User-specific access controls

System Workflow

• Source: Kaggle loan default dataset.
• Size: 32,581 rows, 12 columns.
• Features: person_age, person_income, person_home_ownership, person_emp_length, loan_intent, loan_grade, loan_amnt, loan_int_rate, loan_status, loan_percent_income, cb_person_default_on_file, cb_preson_cred_hist_length.
• Key columns: Include person_age, person_income, loan_amnt, loan_int_rate, loan_intent, and the target: loan_status.

Technical Implementation

Technologies Used

• Django
• PostgreSQL
• Python
• Bootstrap
• Authentication & Authorization

Key Components

Document Processing

• Django
• PostgreSQL
• Python

Database Layer

• PDF upload
• Data extraction
• Auto Categorization
• Validation

Reporting Layer

• Filtering
• Transaction views
• Report generation

Security & Compliance

Security Considerations

• User authentication
• User-specific data access
• Secure document handling
• POPIA-conscious development approach

Challenges & Solutions.

Challenge: Bank Statement Variations

Different banks generate statements using different formats.

Solution

Implemented pattern recognition using regular expressions to support extraction from targeted statement formats.

Challenge: Limited Training Data

Client data could not be used during development.

Solution

Created a development dataset using personal statements while maintaining confidentiality.

Challenge: Categorizing unseen data

Unseen data could not be categorized

Solution

Unknow transactions set to general transactions and allow users to manually categorize same descriptions in bulk.

Results & Impact

Outcomes

• Reduced manual transaction capture effort
• Improved accessibility of financial records
• Enabled faster transaction search and retrieval
• Created a foundation for future automated categorisation
• Improved consistency of financial reporting workflows

Future Roadmap

Planned Enhancements

• Support for multiple bank formats
• Automated categorisation models
• Financial dashboards
• SaaS deployment