Information systems for criminal network analysis: from data representation to decision support under uncertainty

Criminal network analysis seeks to extract actionable intelligence from relational data that are inherently incomplete, uncertain, and heterogeneous. Although a large body of work has developed computational techniques for analyzing such networks, existing surveys remain fragmented across methodological traditions and largely descriptive in nature. From an Information Systems (IS) perspective, this fragmentation obscures a central question: under what information conditions can analytical outputs be considered valid and used for decision-making. This study addresses that gap by reframing criminal network analysis as an IS problem of information evaluation, validity assessment, and decision use under uncertainty. It introduces a theory-anchored evaluative taxonomy structured across three tightly coupled layers—Data Representation, Analytical Processing, and Knowledge Integration—and extends this structure through complementary facets that connect analytical outputs to investigative interpretation and to the minimum data conditions under which such interpretations remain defensible. Beyond the taxonomy itself, the study proposes three explicit operational frameworks: Operationalizing Representation as a Determinant of Investigative Validity, Operationalizing Analytical Validity, and Operationalizing Validity-Preserving Intelligence Integration. Together, these frameworks make explicit how information-quality constraints propagate across the criminal network analysis pipeline, how analytical credibility should be assessed under uncertainty, and how integrated intelligence should be calibrated for responsible investigative use. To substantiate this evaluative logic, the study experimentally examines representative techniques from the Data Representation, Analytical Processing, and Knowledge Integration layers using unified validity-oriented criteria aligned with the proposed frameworks. The results show that methods differ not only in analytical strength, but also in evidentiary robustness, uncertainty tolerance, provenance support, and decision-use suitability. This study clarifies how criminal network analytics should be interpreted, validated, and operationalized in real-world investigative environments.