Abstract:
|
Business failure prediction has been a classic problem in the business community and a wide variety of data mining techniques have become the new favorites of researchers in the field for their outstanding prediction abilities. While majority of the researches focused on profitable corporations, few attempted to study the charitable sector, especially in Canada, partially due to the lack of available data of this sector. However, this study attempts to apply the existing data mining methodologies on a business failure prediction task aiming at Canadian charitable organizations. Efforts are made to do the prediction with Decision Tree (DT) and Random Forest (RT) methods by a comprehensive use of the panel data containing multiple explanatory variables recorded for thousands of charitable organizations during a three-year period, given that these two techniques are not originally designed for data sets with such complex dimensions. A novel approach to handle time series attributes using DT and RF by measuring the Dynamic Time Warping (DTW) dissimilarities between observations is presented.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.