Paper Type

ERF

Abstract

Financial fraud involves deceitful manipulation of financial records, posing significant challenges for auditors and regulators. Despite recent efforts to automate financial audits using predictive analytics, deep learning models lack transparency, deterring effective fraud detection. To address this, the implementation of two widely used explainable models, LIME and SHAP, is proposed to assess their effectiveness in fraud detection. LIME constructs interpretable models around predictions, while SHAP assigns importance values to features, enhancing transparency. This research aims to improve understanding of deep learning techniques in fraud detection, offering insights to regulators, investors, and auditors. By implementing interpretable models, the study aims to produce valuable findings to bolster confidence in predictions generated by deep learning models.

Paper Number

1635

Author Connect URL

https://authorconnect.aisnet.org/conferences/AMCIS2024/papers/1635

Comments

SIGASYS

Author Connect Link

Share

COinS
 
Aug 16th, 12:00 AM

Evaluating Interpretable Models for Financial Fraud Detection

Financial fraud involves deceitful manipulation of financial records, posing significant challenges for auditors and regulators. Despite recent efforts to automate financial audits using predictive analytics, deep learning models lack transparency, deterring effective fraud detection. To address this, the implementation of two widely used explainable models, LIME and SHAP, is proposed to assess their effectiveness in fraud detection. LIME constructs interpretable models around predictions, while SHAP assigns importance values to features, enhancing transparency. This research aims to improve understanding of deep learning techniques in fraud detection, offering insights to regulators, investors, and auditors. By implementing interpretable models, the study aims to produce valuable findings to bolster confidence in predictions generated by deep learning models.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.