Banking System Incidents Analysis Using Knowledge Graph

Abstract

Risk incidents in the banks' systems have caused significant social impacts and economic losses. This study proposes a risk incident knowledge modeling and analysis approach based on the knowledge graphs to realize the effective integration and continuous accumulation of incident knowledge. The authors are the first to analyze the advantages of knowledge graphs in risk incident knowledge integration for the bank's core system. Moreover, they study and compare the related field's state-of-the-art models (including CRF, BiLSTM, BiLSTM-CRF, BERT-BiLSTM-CRF). This paper proposes an improved Bert-BiLSTM-CRF model to perform entity recognition which replaces “individual word mask and training” with “full word mask and training” targeted to solve the problem of low accuracy in the extraction of incident text entities in the banking system. Experiments on 1000 banking system incident material show that the improved Bert-BiLSTM-CRF model outperforms the state-of-the-art models based on the comparison of recall (R), precision (P), and F1-measure, with a 2%-9% improvement in the F1-measure.