The KOJAK Group Finder: Connecting the Dots via Integrated Knowledge-Based and Statistical Reasoning

Abstract

Link discovery is a new challenge in data mining whose primary concerns are to identify strong links and discover hidden relationships among entities and organizations based on low-level, incomplete and noisy evidence data. To address this challenge, we are developing a hybrid link discovery system called KOJAK that combines state-of-theart knowledge representation and reasoning (KR&R) technology with statistical clustering and analysis techniques from the area of data mining. In this paper we report on the architecture and technology of its first fully completed module called the KOJAK Group Finder. The Group Finder is capable of finding hidden groups and group members in large evidence databases. Our group finding approach addresses a variety of important LD challenges, such as being able to exploit heterogeneous and structurally rich evidence, handling the connectivity curse, noise and corruption as well as the capability to scale up to very large, realistic data sets. The first version of the KOJAK Group Finder has been successfully tested and evaluated on a variety of synthetic datasets.

Discovering and Explaining Abnormal Nodes in Semantic Graphs

“…Therefore, finding important or strong connections to a seed suspect is likely to turn up other suspects ("guilt by association"), which is a phenomenon also exploited by some group detection algorithms [18]. The result reveals an interesting phenomenon where suspects have both important and abnormal connection with other suspects, since they can be detected with both types of mechanisms.…”

“…For semantic graphs, Adibi et al [18] describe a method combining a knowledge-based system with MI analysis to identify threat groups given a set of seeds. Krebs [5] applied SNA to the 9/11 terrorists network and suggested that to identify covert individuals; it is preferable to utilize multiple types of relational information to uncover the hidden connections in evidence.…”

Building Knowledge Maps of Web Graphs

“…[4] address a variant of this problem: given an edge-weighted undirected graph, two vertices s, t, and an integer k, find a connected subgraph H of size k containing s, t that maximizes a given goodness function. Other approaches have been proposed to discover groups of persons (e.g., [1]) or simplify networks (e.g., [12]). However, these approaches do not provide algebras to manipulate the objects that are produced.…”

A Generalized Approach for Social Network Integration and Analysis with Privacy Preservation

“…Recent development in link mining [22] of social networks focuses on object ranking [6,32,54], object classification [9,25,36,43,52], group detection [1,33,34,51,65,67], entity resolution [6,7,15], link prediction [13,37,38,40,53], subgraph discovery [31,35,75], graph classification [20,21], and graph generative models [20,68]. Object ranking [6,32,54] utilizes the link structure of a network to prioritize the objects which are represented as nodes in a network.…”