B-LSTM-NB BASED COMPOSITE SEQUENCE LEARNING MODEL FOR DETECTING FRAUDULENT FINANCIAL ACTIVITIES

Deep Learning (DL) in finance is widely regarded as one of the pillars of financial services sectors since it performs crucial functions such as transaction processing and computation, risk assessment, and even behavior prediction. As a subset of data science, DL can learn and develop from their experience, which does not require constant human interference and programming, implying that the technology will improve quickly. By loading an Ensemble Model (EM), a Deep Sequential Learning (DSL)model, and additional upper-layer EM classifier in the correct order, a new Contained-In-Between (C-I-B) composite structured DSL model is recommended in this article. In cases like Fraud Detection System (FDS), where the data flow comprises vectors with complex interconnected characteristics, DL models with this structure have proven to be highly efficient. Finally, by utilizing optimized transaction eigenvectors, a NB classifier is trained. This strategy is more effective than most standard approaches in identifying transaction fraud. The proposed model is evaluated for its accuracy, Recall and F-score, and the results show that the model has better performance against its counterparts.

Automated summary

Deep Learning in finance plays a crucial role in transaction processing, risk assessment, and behavior prediction. This article proposes a new composite structured Deep Sequential Learning model for complex data flows, which has proven to be highly efficient in cases like Fraud Detection Systems. Additionally, a trained NB classifier utilizing optimized transaction eigenvectors outperforms standard approaches in identifying transaction fraud.