PresentationPDF Available

New supervised classification approach as Networked Pattern Recognition Framework

Authors:

Abstract and Figures

This aim of this study is to propose a new classification framework as Networked Pattern Recognition (NEPAR) for different classification problems. In most research studies, classification focuses on either individual observations, which do not consider the dynamic interactions, which ignores the functional roles of observations. When they capturing interactions, they just give a general idea about networks. In this study, we propose a unified approach that combines pattern classification techniques and dynamic interactions for better classification approach. Therefore, the NEPAR and five different classification methods (SVM, NB, LR, DT, and kNN) are developed by adding information from the proposed networks (as seen in Figure 2-3). Figure 1. Combining network metrics and pattern recognition. As seen in Figure 1, information from observations is extracted by building the network, and feature properties for each observation are used to classify the output. For the results, we compare three approaches: (1) classic approach that uses traditional pattern recognition techniques; (2) the networked approach that uses pattern recognition techniques on the network topology; and (3) the unified approach that combines network topology and real data with pattern recognition methods (see Figure 1). Figure 2. Networks for the Pima Indian diabetes dataset and Australian credit card approval dataset. More specifically, a new weighted heterogeneous similarity function is also proposed to estimate relationships among interactive events. In the second phase of the framework, combining pattern-recognition techniques with network-based approaches. In this research, we propose a new hybrid detection framework in the proposed network topology [1].
Content may be subject to copyright.
New supervised classification approach as Networked Pattern Recognition Framework
Salih Tutun1*, Mohammad T Khasawneh1
1Systems Science and Industrial Engineering, Binghamton University, Binghamton, NY, 13850, USA
*stutun1@binghamton.edu
This aim of this study is to propose a new classification framework as Networked Pattern Recognition (NEPAR) for
different classification problems. In most research studies, classification focuses on either individual observations,
which do not consider the dynamic interactions, which ignores the functional roles of observations. When they
capturing interactions, they just give a general idea about networks. In this study, we propose a unified approach that
combines pattern classification techniques and dynamic interactions for better classification approach. Therefore, the
NEPAR and five different classification methods (SVM, NB, LR, DT, and kNN) are developed by adding
information from the proposed networks (as seen in Figure 2-3).
Figure 1. Combining network metrics and pattern recognition.
As seen in Figure 1, information from observations is extracted by building the network, and feature properties for
each observation are used to classify the output. For the results, we compare three approaches: (1) classic approach
that uses traditional pattern recognition techniques; (2) the networked approach that uses pattern recognition
techniques on the network topology; and (3) the unified approach that combines network topology and real data with
pattern recognition methods (see Figure 1).
Figure 2. Networks for the Pima Indian diabetes dataset and Australian credit card approval dataset.
More specifically, a new weighted heterogeneous similarity function is also proposed to estimate relationships
among interactive events. In the second phase of the framework, combining pattern-recognition techniques with
network-based approaches. In this research, we propose a new hybrid detection framework in the proposed network
topology [1].
Figure 3. Networks for the breast cancer dataset and the StarPlus fMRI dataset.
As a result, the networks (as seen in Figure 2 and Figure 3) are built to see how the events are similar and how they
interact with each other [2]. Based on the network metrics such as degree centrality, closeness centrality,
betweenness centrality, in-degree centrality, out-degree centrality, load centrality and harmonic centrality, the
pattern recognition techniques are applied to detect the credit card approval, breast cancer diagnosing, schizophrenia
disease in fMRI, and diabetic disease [3]. In conclusion, the proposed approach was tested and validated using real-
world case studies. Five different datasets (Australian credit approval, diabetes, breast cancer, SturPlus fMRI,
German credit) are used to show the framework outperforms other traditional classification.
1. Tutun, Salih, Mohammad T. Khasawneh, and Jun Zhuang. New framework that uses patterns and
relations to understand terrorist behaviors, 2017 78 Expert Systems with Applications 358-375.
2. Chenoweth, E., & Lowham, E. (2007). On classifying terrorism: A potential contribution of cluster
analysis for academics and policy-makers. Defence & Security Analysis, Vol:23 , 345-357.
3. Chen, H., Dark web: Exploring and data mining the dark side of the web, 2011 Vol:30. Springer Science &
Business Media.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Terrorism is defined as a premeditated, politically motivated violence perpetrated against noncombatant targets by subnational groups or clandestine agents, usually intended to influence an audience. There are alternative ways to conceive terrorist typologies or the classification of terrorist groups for analysis and response. Cluster analysis provides a technique for large scale comparisons while maintaining the contextuality and comprehensiveness of individual incidents. There are two critical choices in setting up a cluster analysis: choice of the measure of similarity within the data and choice of the algorithm to determine groupings. The analysis is run on 259 incidents using a Jaccard coefficient as a measure of similarity and an average between groups linkage as the computational algorithm. Ten core cluster have been identified which were classified under the bombing and the non-bombing clusters. For the former: bombings of a public population where a liberation group takes responsibility; bombings of a public population at a commercial target where groups take responsibility; bombings of a public population at a commercial target by an unknown groups; bombings of official population at official targets by unknown groups; and the bombings of foreign populations at military targets where a group takes responsibility. For the latter: gun attacks where a righteous vengeance group takes responsibility; assassination of foreign population with guns by unknown groups; attacks on foreign, official populations in open air targets where groups take responsibility; attacks on official populations at official targets with no deaths where a group takes responsibility; and kidnappings at open-air targets with small casualties and no deaths. Overall, terrorist groups should thus be classified not only on the basis of their motives, nationalities, and religions, but also on the basis of their tactics, destructiveness, and targets.
Conference Paper
This talk will review the emerging research in Terrorism Informatics based on a web mining perspective. Recent progress in the internationally renowned Dark Web project will be reviewed, including: deep/dark web spidering (web sites, forums, Youtube, virtual worlds), web metrics analysis, dark network analysis, web-based authorship analysis, and sentiment and affect analysis for terrorism tracking. In collaboration with selected international terrorism research centers and intelligence agencies, the Dark Web project has generated one of the largest databases in the world about extremist/terrorist-generated Internet contents (web sites, forums, blogs, and multimedia documents). Dark Web research has received significant international press coverage, including: Associated Press, USA Today, The Economist, NSF Press, Washington Post, Fox News, BBC, PBS, Business Week, Discover magazine, WIRED magazine, Government Computing Week, Second German TV (ZDF), Toronto Star, and Arizona Daily Star, among others. For more Dark Web project information, please see: http://ai.eller.arizona.edu/research/terror/ .