Title page for etd-0023117-230904


[Back to Results | New Search]

URN etd-0023117-230904
Author Yu-Feng Hsu
Author's Email Address No Public.
Statistics This thesis had been viewed 5213 times. Download 0 times.
Department Information Management
Year 2016
Semester 1
Degree Ph.D.
Type of Document
Language English
Title The Evaluation and Prediction of the Going-Concern Status for Companies: A Model Based on Structured and Un-Structured Data
Date of Defense 2017-01-06
Page Count 187
Keyword
  • Going-concern prediction
  • Ensemble framework
  • Text mining
  • Random forest
  • Financial news articles
  • Abstract Ascertainment of the going-concern status of a company is a critical issue for investors and stockholders. In the Accounting and Auditing domain, the going-concern is a well-known concept used to measure whether a company has the resources to operate indefinitely or not. However, it is difficult to evaluate a company’s financial condition in today’s complicated business environment. To make this easier, some researchers have proposed new methods to assist in the auditing process. The majority of these studies have proposed single models, applying numerical data gathered from financial statements to verify their methodology. However, shortcomings remain such as a lack of flexibility, generalizability and time efficiency. In order to address these issues, in this study, we introduce a framework called the ensemble method and adopt financial news as source of data source. One of the characteristics of the ensemble framework is that a weaker algorithm can be easily replaced by another if it is better. In addition, financial news is an important source of information, especially given the issue of the lack of annual reports for a new to market company. Text mining techniques are applied to capture messages hidden in financial news, and convert the textual data to a numerical format for implementation in the experiments. In study one, the random forest method is applied to implement the concepts of the ensemble method. The experimental results show that the random forest method outperforms the baseline methods in terms of accuracy rate, ROC area, kappa value, type II error, precision and recall rate. In addition, the experimental results obtained in study two reflect that text mining techniques perform well for going-concern prediction. Financial news is a useful data source for analyzing the going-concern status of a company before the issue of an annual report or for a new to market company, where such reports do not yet exist.
    Advisory Committee
  • Chia-Mei Chen - chair
  • Keng-Pei Lin - co-chair
  • Wann-cherng Wang - co-chair
  • Wei-Po Lee - advisor
  • Bing-Chiang Jeng - advisor
  • Files
  • etd-0023117-230904.pdf
  • Indicate in-campus at 5 year and off-campus access at 5 year.
    Date of Submission 2017-01-24

    [Back to Results | New Search]


    Browse | Search All Available ETDs

    If you have more questions or technical problems, please contact eThesys