Open vs. Close Source Decision Tree Algorithms: Comparing Performance Measures of Accuracy, Sensitivity and Specificity
Document Type
Conference Proceeding
Publication Date
2017
Publication Title
Proceedings of the CONISAR
ISSN
2167-1508
Abstract
Data Science research is trending due the abundance of publicly available data and open source and close (proprietary) tools available. Currently, an abundant amount of research exists on various data science techniques, tools and mining of medical data and big data. However, there is little to nonexistent research, which actually compares closed and open source algorithms. This research compared a closed source algorithm (Microsoft Decision Tree ) with open source algorithms (CART and C4.5) performances for accuracy, sensitivity, and specificity using data form the U.S. government’s Surveillance, Epidemiology, and End Results Program (SEERS). Data was downloaded, converted from raw data to structured data using a custom designed python script and transformed via the removal of missing and irrelevant data, and outliers. Predictive modeling results for accuracy, sensitivity, and specificity, indicated that closed algorithms have the best accuracy and specificity.
Recommended Citation
Khan, Sushmita, Hayden Wimmer, Loreen Marie Powell.
2017.
"Open vs. Close Source Decision Tree Algorithms: Comparing Performance Measures of Accuracy, Sensitivity and Specificity."
Proceedings of the CONISAR, 10.
source: http://proc.conisar.org/2017/pdf/4504.pdf
https://digitalcommons.georgiasouthern.edu/information-tech-facpubs/75
