Identification and Analysis of Financial Technology Risk Factors Based on Textual Risk Disclosures

Abstract

With the development of financial technology (referred to as fintech), the risks faced by fintech companies have received increasing attention. This paper uses the Sentence Latent Dirichlet Allocation (Sent-LDA) topic model to comprehensively identify risk factors in the fintech industry based on textual risk factors disclosed in Form 10-K. Furthermore, this paper analyzes the importance of risk factors and the similarities of the risk factors for the whole fintech industry and different fintech sub-sectors from the perspectives of risk factor types and risk factor contents. In the empirical analysis, 53,452 risk factor headings of 34 fintech companies included in the KBW Nasdaq Financial Technology Index (KFTX) over the period 2015–2019 are collected. The empirical results show that 20 risk factors of the fintech industry are identified. However, the important risk factors vary differently among different fintech sub-sectors. For the analysis of risk factor similarity, mean values of similarity of risk factor types and the similarity of risk factor contents both increased from 2015 to 2019, which indicates that the risks faced by fintech companies are becoming increasingly similar. The mean value of similarity of risk factor contents is 42.13%, while the mean value of similarity of risk factor types is 80.93%. Thus, although the types of risk factors faced by different fintech companies are similar, the contents of risk factors disclosed by different companies are still quite different. The comprehensive identification of fintech risk factors lays an important foundation for the further measurement and management of risks in the fintech industry. In the feature, we will further make effective risk estimations of the fintech industry based on the identified fintech risk factors.

Machine learning in internet financial risk management: A systematic literature review

“…The research on the topic of internet financial risk has a long history [ 21 ], encompassing both quantitative empirical analyses [ 22 ] and qualitative descriptions [ 1 ], as well as comprehensive review studies [ 13 ]. There are analyses employing quantitative platform data [ 2 , 14 ] and those conducted using textual data [ 23 , 24 ]. Studies have delved into various subtopics such as risk perception [ 22 ], risk identification [ 24 ], and risk regulation [ 12 ], rendering the research on internet financial risk quite extensive.…”

“…There are analyses employing quantitative platform data [ 2 , 14 ] and those conducted using textual data [ 23 , 24 ]. Studies have delved into various subtopics such as risk perception [ 22 ], risk identification [ 24 ], and risk regulation [ 12 ], rendering the research on internet financial risk quite extensive.…”

Random forest model in tax risk identification of real estate enterprise income tax

“…Effectively identifying high-risk tax-paying enterprises and tax-paying risk points, the random forest algorithm empowers tax authorities to efficiently and accurately categorize and manage taxpayers with varying risk levels. Comparative to Wei et al’s (2022) tax risk identification model, the model proposed in this paper excels in stability and accuracy [ 32 ]. Furthermore, when juxtaposed with Zhao’s (2022) tax risk identification model, the optimized model exhibits a heightened ability to capture key risk factors [ 33 ].…”