DETECTION VIOLATION WITH INFORMATION RETRIEVAL IN FINANCIAL ADVERTISEMENTS USING DISTILBERT AND RANDOM FOREST IN INDONESIA

Abstrak

The evolving financial market in Indonesia has introduced significant challenges in regulatory compliance monitoring, especially with the increase in complex financial advertisements. Traditional rule-based and manual compliance methods struggle with scalability, accuracy, and adaptability to diverse advertising formats. This research addresses these challenges by developing a compliance monitoring system utilizing DistilBERT, a lean version of BERT, to create dense text embeddings, and Random Forest, known for handling high-dimensional data, to classify advertisement compliance. Through an Indonesian-specific dataset of financial advertisements, the system identifies non-compliant content effectively, enhancing both accuracy and efficiency in monitoring. This hybrid approach contributes a scalable and adaptable solution that aligns with Indonesia's regulatory landscape, ensuring that financial advertisements meet legal standards.