https://repository.kcau.ac.ke/handle/123456789/713 2026-04-20T00:48:37Z 2026-04-20T00:48:37Z Arusei, Mike K. Dr. Njenga, Stephen https://repository.kcau.ac.ke/handle/123456789/1601 2026-01-17T00:00:48Z 2025-01-01T00:00:00Z

Detecting Data Exfiltration Anomalies in Academic Networks Using the Isolation Forest Algorithm Arusei, Mike K.; Dr. Njenga, Stephen Academic networks face increased risks of data exfiltration due to sensitive personal information and research data. Traditional supervised detection models rely on labeled datasets which are often unavailable in resource constrained institutions. This study investigates the applicability of the unsupervised Isolation Forest algorithm for detecting anomalous network traffic indicative of data exfiltration. The research utilized the CICIDS2017 dataset focusing on the Thursday-Working Hours-Afternoon-Infiltration subset. Key features including Flow Duration, Total Fwd Packets, Flow Bytes/s, Flow IAT Mean, and Destination Port were preprocessed and normalized for modeling. The model achieved a precision of 1.00, recall of 0.99 and F1-score of 1.00 for anomalous traffic detection successfully identifying approximately 4.8% of flows as anomalous. Comparative analysis with previous methods, including supervised Random Forest and SVM demonstrated that Isolation Forest offers competitive accuracy with lower computational overhead and does not require labeled data. The findings highlight the algorithm’s suitability for academic network monitoring, providing an effective early warning mechanism while emphasizing the importance of threshold tuning to reduce false positives.

2025-01-01T00:00:00Z Mwendia, Simon N Mburu, Lucy W Mpia, Héritier N https://repository.kcau.ac.ke/handle/123456789/1409 2023-07-05T09:02:33Z 2022-01-01T00:00:00Z

Predicting Employability of Congolese Information Technology Graduates Using Contextual Factors: Towards Sustainable Employability Mwendia, Simon N; Mburu, Lucy W; Mpia, Héritier N Predicting employability in an unstable developing country requires the use of contextual factors as predictors and a suitable machine learning model capable of generalization. This study has discovered that parental financial stability, sociopolitical, relationship, academic, and strategic factors are the factors that can contextually predict the employability of information technology (IT) graduates in the democratic republic of Congo (DRC). A deep stacking predictive model was constructed using five different multilayer perceptron (MLP) sub models. The deep stacking model measured good performance (80% accuracy, 0.81 precision, 0.80 recall, 0.77 f1-score). All the individual models could not reach these performances with all the evaluation metrics used. Therefore, deep stacking was revealed to be the most suitable method for building a generalizable model to predict employability of IT graduates in the DRC. The authors estimate that the discovery of these contextual factors that predict IT graduates’ employability will help the DRC and other similar governments to develop strategies that mitigate unemployment, an important milestone to achievement of target 8.6 of the sustainable development goals.

2022-01-01T00:00:00Z Mburu, Lucy W Mwendia, Simon N Mpia, Héritier N https://repository.kcau.ac.ke/handle/123456789/1402 2023-07-04T09:30:17Z 2023-01-01T00:00:00Z

Applying Data Mining in Graduates’ Employability Mburu, Lucy W; Mwendia, Simon N; Mpia, Héritier N Envisaging an adequate IT/IS solution that can mitigate the employability problems is imperative because nowadays there is a high rate of unemployed graduates. Thus, the main goal of this systematic literature review (SLR) was to explore the application of data mining techniques in modeling employability and see how those techniques have been applied and which factors/variables have been retained to be the most predictors or/and prescribers of employability. Data mining techniques have shown the ability to serve as decision support tools in predicting and even prescribing employability. The review determined and analyzed the machine learning algorithms used in data mining to either predict or prescribe employability. This review used the PRISMA method to determine which studies from the existing literature to include as items for this SLR. Hence, 20 relevant studies, 16 of which are predicting employability and 4 of which are prescribing employability. These studies were selected from reliable databases: ScienceDirect, Springer, Wiley, IEEE Xplore, and Taylor and Francis. According to the results of this study, various data mining techniques can be used to predict and/or to prescribe employability. Furthermore, the variables/factors that predict and prescribe employability vary by country and the type of prediction or prescription conducted research. Nevertheless, all previous studies have relied more on skill as the main factor that predict and/or prescribe employability in developed countries and none studies have been conducted in unstable developing countries. Therefore, the need to conduct research on predicting or prescribing employability in such countries by trying to use contextual factors beyond skill as features.

2023-01-01T00:00:00Z Ngaii, John W Manene, Moses Njui, Francis https://repository.kcau.ac.ke/handle/123456789/1401 2023-07-04T09:20:03Z 2020-01-01T00:00:00Z

Further Construction Of Balanced Arrays Ngaii, John W; Manene, Moses; Njui, Francis The relation between balanced arrays and two other combinatorial structures, namely, orthogonal arrays and transitive arrays is pointed out. We provide three new and simple but rather stringent methods of constructing balanced arrays of any strength provided that the balanced arrays exist. A theorem that enables one to generate a balanced array from several known balanced arrays has been proved. The existence results of some types of balanced arrays based on the existence of some types of Hadamard matrices have also been proved.

2020-01-01T00:00:00Z