Start Date
14-12-2012 12:00 AM
Description
Company movements often are headlines of the press, helping managers to gauge the risk factors. While most corporate analyses are based on numerical financial figures stated in corporate reports, relatively little work has been done to reveal company risk factors from news articles. In this research, we developed an integrated framework for automatic assessment of company risks from news articles. We present a study of using the framework to categorize four IT companies' risk factors. Our experimental findings show that the three chosen classification techniques - Support Vector Machine, Naïve Bayes, and Logistic Regression - achieved encouraging results. NB outperformed both SVM and LR, while LR outperformed SVM in terms of precision, recall, and F-measure. This research addresses an important concern of risk management that received relatively less attention from previous works. The results demonstrate a strong potential for industry deployment.
Recommended Citation
Chung, Wingyan and Zhu, Min, "Risk Assessment Based on News Articles: An Experiment on IT Companies" (2012). ICIS 2012 Proceedings. 94.
https://aisel.aisnet.org/icis2012/proceedings/ResearchInProgress/94
Risk Assessment Based on News Articles: An Experiment on IT Companies
Company movements often are headlines of the press, helping managers to gauge the risk factors. While most corporate analyses are based on numerical financial figures stated in corporate reports, relatively little work has been done to reveal company risk factors from news articles. In this research, we developed an integrated framework for automatic assessment of company risks from news articles. We present a study of using the framework to categorize four IT companies' risk factors. Our experimental findings show that the three chosen classification techniques - Support Vector Machine, Naïve Bayes, and Logistic Regression - achieved encouraging results. NB outperformed both SVM and LR, while LR outperformed SVM in terms of precision, recall, and F-measure. This research addresses an important concern of risk management that received relatively less attention from previous works. The results demonstrate a strong potential for industry deployment.