A methodological framework for integration of data mining in organizations
abstract: In this thesis, we address the problems and challenges faced by data mining (DM) practitioners in the initial stages of DM technology integration in organizations. The findings (with minor modifications) may also be applied in other analytics domains, i.e. forecasting/extrapolation, modeling, experimental design, simulation, and optimization.
While it is evident that DM now represents a significant technology for strategic applications, there appears to be a dearth of empirical studies that consider in detail the initial (embryonic) stages in DM management to enable an appropriate foundation for its later successful implementation. Most extant theory either fails to consider the distinctive context and aim of the embryonic DM process or focuses on large-scale DM implementation. Yet, in the great majority of organizations, the embryonic DM process is a sine qua non of enterprise-wide DM integration.
Our research therefore aimed to propose a methodological framework ā€“ a system of principles, practices, and procedures ā€“ to guide practitioner decision making through the embryonic DM process. We hypothesize that the application of the methodological framework increases the likelihood of success of embryonic DM. Due to the nature of the artifact, we applied a design science research methodology. Embedded within the design process we also applied a structured-case framework to identify best practices of embryonic DM. Primary data was principally collected through semi-structured interviews with DM practitioners. The proposed formulation of a methodological framework was validated and reported through a series of case studies.
Our findings indicate a significant range of considerations and reveal additional issues for applied decision making in the context of DM requirements and process success. Addressing best practices of embryonic DM a critical success factors framework is proposed. It suggests four success measures and seven success factors which, if managed well, lead to success. Moreover, a process model for carrying out embryonic DM is designed. The findings extend extant theory on DM implementation and can therefore be used for comparative studies and the development of cumulative knowledge.