Serkan Üçer1, Yunuscan Koçak2, Tansel Ozyer3, Reda Alhajj4. 1. Department of Computer Engineering, TOBB University, Ankara, Turkey; Department of Computer Science, University of Calgary, Calgary, Alberta, Canada. Electronic address: sucer@etu.edu.tr. 2. Department of Computer Engineering, TOBB University, Ankara, Turkey; Department of Computer Science, University of Calgary, Calgary, Alberta, Canada. Electronic address: y.kocak@etu.edu.tr. 3. Department of Computer Engineering, TOBB University, Ankara, Turkey; Department of Computer Science, University of Calgary, Calgary, Alberta, Canada. Electronic address: ozyer@etu.edu.tr. 4. Department of Computer Engineering, TOBB University, Ankara, Turkey; Department of Computer Science, University of Calgary, Calgary, Alberta, Canada. Electronic address: alhajj@ucalgary.ca.
Abstract
BACKGROUND AND OBJECTIVES: Social Network Analysis is an attractive approach to model and analyze complex networks. In recent years, several bioinformatics related networks have been modeled and analyzed thoroughly using social network analysis. The objective of this study is to build a social network analysis based classifier for time sequential data. METHODS: In this work, we model a genomic time sequential data as a 'social' network of interactions. We define interactions as similarity of patients' measurements. Using this 'genomic social network', we develop a classification model called Social Network Analysis-based Classifier. RESULTS: We conducted some experiments to demonstrate how the developed Social Network Analysis-based Classifier outperforms traditional classifiers by effectively classifying a time sequential genomic dataset. Best achieved accuracy is 64.51% and best f-measure is 78.34%. CONCLUSIONS: Our study emphasized Social Network Analysis-based Classifier Model as a powerful technique for analyzing a time sequential dataset. Eventually, the plan is to develop and evolve the Social Network Analysis-based Classifier model into a general classifier.
BACKGROUND AND OBJECTIVES: Social Network Analysis is an attractive approach to model and analyze complex networks. In recent years, several bioinformatics related networks have been modeled and analyzed thoroughly using social network analysis. The objective of this study is to build a social network analysis based classifier for time sequential data. METHODS: In this work, we model a genomic time sequential data as a 'social' network of interactions. We define interactions as similarity of patients' measurements. Using this 'genomic social network', we develop a classification model called Social Network Analysis-based Classifier. RESULTS: We conducted some experiments to demonstrate how the developed Social Network Analysis-based Classifier outperforms traditional classifiers by effectively classifying a time sequential genomic dataset. Best achieved accuracy is 64.51% and best f-measure is 78.34%. CONCLUSIONS: Our study emphasized Social Network Analysis-based Classifier Model as a powerful technique for analyzing a time sequential dataset. Eventually, the plan is to develop and evolve the Social Network Analysis-based Classifier model into a general classifier.
Authors: Md Zakir Hossain; Elena Daskalaki; Anne Brüstle; Jane Desborough; Christian J Lueck; Hanna Suominen Journal: BMC Med Inform Decis Mak Date: 2022-09-15 Impact factor: 3.298