K. Niki Kunene

An approach for predicting and describing patient outcome using multicriteria decision analysis and decision rules

Patient outcome in brain trauma patients is affected by a multiplicity of factors, beginning with ambulatory transportation and routing, to the grade of the receiving facility and treatment therein, and finally the treatment and monitoring in definitive care (the brain trauma intensive care unit). Factors and events in each of these phases can be modeled as a multicriteria problem, where the objective is to optimize patient outcome; moreover, a more comprehensive model can embody the interactions of all three phases. This study focuses on modeling the factors that affect patient outcome in definitive care and on expressing these in machine readable format so that we can better describe or predict patient outcome using data mining tools. We use multicriteria decision analysis and decision rules for knowledge representation. Preliminary results suggest that the incorporation of a priori knowledge does help better predict or describe patient outcome when using decision tree induction.

Frugal information systems IS

A discussion of the factors contributing to a likely increase in austerity establishes the case for a greater emphasis on frugal information systems (IS), which is defined as an information system that is developed and deployed with minimal resources to meet the pre-eminent goal of the client. The U-constructs are adopted as a foundation for a frugal IS, and two simple messaging system-based case studies illustrate their application. Design guidelines and practices for frugal IS are considered. The challenges and potential gains of frugal IS are reviewed.

ICT Capacity as the Investment and Use of ICT: Exploring its Antecedents in Africa

In this paper, we investigate the extent to which economic, human development endowment, political, and population conditions influence information and communication technology (ICT) capacity in African nations using ordinary least squares. We use five variables to measure ICT capacity: ICT expenditure as a percent of GDP (ICTE), ICT expenditure per capita in US dollars (ICTEP), investment in telecoms with private participation (INVTE), Internet users per 100 people (INTUS), and mobile phone subscribers per 1000 people (MOBPS). Unlike ICT adoption and diffusion that focuses on the use of ICT, ICT capacity measures public and private investments in ICT and the use of ICT. The results indicate that ICT capacity of a nation increases with increased (high) corruption index. Further, ICT investment by the private sector depends mainly on human development factors and is not influenced by the perceptions of corruption about a nation. We present theoretical and practical implications of the findings for developing nations in general and African nations in particular.

Architecture for Managing Knowledge on Cybersecurity in Sub-Saharan Africa

In this paper, we present architecture for managing knowledge on cybersecurity in Sub-Saharan Africa. This architecture captures the critical contextual factors in the Sub-Saharan African region. The architecture enables the creation, storage/retrieval, transfer and application of knowledge on cybersecurity especially for home users while providing awareness and enforcement mechanisms to help home users protect themselves against cyber threats. The architecture can facilitate the building of expertise on cybersecurity among the stakeholders. The implications of the proposed architecture are presented.

Performance assessment of data mining methods for loan granting decisions: a preliminary study

After the greatest financial debacle since the great depression, the need for accurate and systematic assessment of loan granting decisions has never been more important than now. The paper compares the classification accuracy rates of six models: logistic regression (LR), neural network (NN), radial basis function neural network (RBFNN), support vector machine (SVM), k-Nearest Neighbor (kNN), and decision tree (DT) for loan granting decisions. We build models and test their classification accuracy rates on five very versatile data sets drawn from different loan granting decision contexts. The results from computer simulation constitute a fertile ground for interpretation.