Financial Fraud Detection Using Deep Learning Approach

  • Ajit Kr. Singh Yadav*, Marpe Sora
Keywords: Financial Misstatements, Financial Fraud, Management Discussion and Analysis Bag of Words, Deep Learning,

Abstract

Financial misstatement and fraud detection is very crucial in text mining. The researchers use mostly quantitative data and data mining techniques for financial fraud detection. The textual data and text mining can also be used to detect financial misstatements and frauds effectively. For this reason, we have used MD&A reports for fraud detection purpose. This paper develops two models “Bag of Words” and “Deep learning” approaches to extract text classification features to be analysed by various classification techniques. With “Bag of Words” method the MD&A text is analyzed by Loughran and McDonald word list and gives Sentiment_TM scores ranged from -1 to 1. After data preprocessing the words are categorized in positive or negative word list. The scores are calculated by difference of positive and negative scores. The Deep leaning algorithms mainly automate the extraction of information from the data. Due to limited research in financial statement analysis the aim of this paper is to apply DNN over MD&A and analyze the sentiments for financial fraud detection and evaluate the performance of DNN. With Deep Learning method the MD&A text is analyzed by Alchemy Language API and gives the “Sentiment_DL” scores -1 to 1, where 1 represents positive sentiment, -1 represents negative sentiments and 0 represents neutral sentiments. It also provides the confidence score “JOY” ranged from 0 to 1. Here we are using four classification techniques for the two target variables “misstatement” and “fraud” with the sentiment features extracted by two models. The result shows that deep learning-based sentiment features performs better than others. it is also observed by the results that deep learning approach more effectively detects the fraud than misstatement.

Published
2021-08-03
How to Cite
Marpe Sora, A. K. S. Y. (2021). Financial Fraud Detection Using Deep Learning Approach. Design Engineering, 6254-6267. Retrieved from http://thedesignengineering.com/index.php/DE/article/view/3124
Section
Articles