| dc.description.abstract | This project is aimed at understanding the potential and limitations of modern machine learning algorithms in automatically detecting front companies. We employed predictive classification models to analyse a company's transaction history, with the goal of developing the most effective approach to solve the classification task.
The research was conducted in collaboration with B4 Investigate, a young company specializing in developing software for fraud detection and financial damage mitigation. The dataset used for analysis and model development was provided by B4 Investigate and was analysed using various Python libraries and intrinsic tools.
Front companies have a pervasive presence worldwide. While there might be valid justifications for utilizing a front, in most cases, they are often employed to conceal engagement in illegal or dubious activities. One significant hurdle is the arduous task of identifying such entities, as inadvertent association with a front company can lead to substantial financial or reputational harm.
Through our analysis, several important insights emerged. We discovered the significance of creating suitable representations of relevant variables, such as the country of registration for each company in the dataset as well as conveying the context of a company's regular business activities when employing the predictive solution. The implementation of these measures significantly improved the performance of the prediction task. Additionally, tree-based algorithms appear to be the most suitable for learning the correct indicative patterns in this specific prediction task. The findings suggest that modern machine learning algorithms indeed have the potential to serve as effective tools for automated detection of front companies, underscoring the value of further exploration in this field through future research. This approach holds promise and can provide tangible value to companies and non-profit organizations by accurately identifying any undisclosed associations with front companies, thus mitigating the risks of financial and reputational damage. | en_US |
