Automatic Extraction of Financial Data in Credit Rating Analysis

Sammanfattning: With the increasing use of big data and automatization, financial data extraction is of growing importance in the financial industry. The thesis examines how an extraction system can be developed for extracting relevant data for credit rating analysis. The system is designed to collect financial reports, extract relevant information, and identify failed extractions. Prerequisites were identified by conducting a qualitative literature study, as well as holding meetings with employees at a credit rating analysis company to align the system’s functionality with the company’s processes. The results showed that an automatic extraction can be implemented. The system was trained through a manual review process, resulting in an increase in performance. Following the training, the system was able to identify and extract all target data correctly. However, in some reports, certain target data was missing and these were treated as failures by the system. In summary, a system that extracts all existing target data was implemented.