School of Technology
Permanent URI for this collectionhttps://repository.kcau.ac.ke/handle/123456789/18
Browse
133 results
Search Results
Item Empirical evaluation of lstm-based stock price forecasting on the Nairobi securities exchange in Kenya(KCA University, 2025) Kitheka, Hellen K.Having a precise prediction of stock prices can be of great importance in investment policies and budgeting. The study used the Long Short-Term Memory (LSTM) neural networks to forecast stock price changes through the formation of sequential, nonlinear relationships in financial time series data. Our data provenance and relevance are enhanced by the fact that the dataset contains the official daily stock prices of the official Nairobi Securities Exchange (NSE). Exploratory data analysis was performed before training the prediction model to carefully preprocess and study the stock prices with performance measured against a baseline with common traditional machine learning algorithms including Random Forest, Support Vector Regression, K-Nearest Neighbors and ARIMA. The LSTM generated the lowest RMSE values and maximum R2 values compared to its baselines. To enhance the interpretability of the model even more, we enhanced transparency with the help of SHAP (SHapley Additive exPlanations) analysis that revealed the most influential features used in predicting stock prices. The findings reflect that deep learning is effective in financial forecasts and give a consistent and versatile structure of comparable predictive modeling in capital markets. However, the analysis was limited by the data and the calculation capabilities, which can affect the generalization to high-frequency trading scenarios. Future work should consider hybrid deep learning architecture, including further macroeconomic and sentiment indicators, and testing the framework in other emerging markets to increase robustness and generalizability.Item A linear regression model of secure remote access to enterprise networks by employees: a case study of Kericho county(KCA University, 2024) Keter, Edwin K.The Covid-19 pandemic has made working from home the usual practice for many companies and organizations around the world. While remote work has provided flexibility and kept things running during unusual times, it has also given criminals new chances to take advantage of weaknesses, causing serious cyber-attacks like scams, fake emails, and hacking. Protecting remote work systems is now more important than ever. This study looks at how people working from home handle security problems, develops a special security plan for remote work, and checks if it works well. The study used a survey to gather information about corporate networks and weaknesses in remote access. Data was collected from teleworkers at different levels within organizations to find out what affects remote access security. The process of collecting and analyzing the data was well-organized, and the results were checked using a statistical method called Ordinary Least Squares (OLS) linear regression. This made sure the findings were reliable and accurate. The research found six important things that affect secure remote access: Technology, Organizational Factors, Employees, Monitoring & Evaluation, Resource Management & Controls, and Data Protection & Monitoring. These six areas together form a guide for companies to improve their information security while allowing employees to work remotely. The OLS linear regression analysis showed that the model can predict remote access security well, with an Adjusted R-squared value of 0.634. This means that the six independent variables explain 63% of the changes in remote access security. ANOVA tests also confirmed that these variables are statistically significant for predicting remote access security. This research is important for organizations that use county and enterprise networks. The study suggests that organizations should use strong security measures, such as secure devices, employee training, ongoing monitoring, and better management of resources. By following these recommendations, organizations can reduce cybersecurity risks and create safe and efficient remote work settings. This research presents a complete and proven model for teleworking security, designed to handle the special problems caused by working from home during the Covid-19 pandemic. The model provides a useful way for companies to protect their computer networks from increasing cyber dangers. By focusing on safe systems and keeping an eye on potential issues, the study helps improve remote work methods and makes sure companies can stay strong against ongoing cybersecurity problems.Item A random forest model to predict malaria outbreaks A case study of Kisii county(KCA University, 2025) Nyabuto, Joyline M.In Sub-Saharan Africa, malaria has remained as a major source of morbidity and mortality, and a constant menace to the population health and socio-economic balance. Seasonal epidemics in highlands areas like Kisii County of Kenya create a huge burden to the healthcare systems, interfere with livelihood, and hamper development in the area. Despite the major advances achieved in reducing malaria rates in the country, the fact that it still is spread in highlands indicates the weakness of the existing control and monitoring measures. Conventional forecasting methods such as logic regression and ARIMA models have been actively used in predicting trends in malaria but their assumptions of linearity, normality, and stationarity makes them less useful to the study of the non-linear and dynamic relationships that exist between environmental, epidemiological, and socio-economic variables that define malaria transmission in highland micro-ecologies. Machine learning, especially ensemble-based ones, is a promising new direction in the development of malaria prediction. Ensemble models can learn more of complex interactions using heterogeneous data, hence able to adjust to changing epidemiological situations and produce more precise and useful information. This paper thus builds and compares an ensemble learning model to forecast malaria outbreaks in Kisii County that incorporates Random Forest (RF), XGBoost and RF-XGBoost hybrid model. The framework uses ecological, climatic, socio-economic, and epidemiological predictors to show that highland malaria occurs in a multifactorial manner. To resolve the important data issues, methodological advances like the Synthetic Minority Over-Sampling Technique (SMOTE), adaptive class weight, and cost-sensitive learning are used to curb the issue of class imbalance, and temporal feature weighting and sliding-window retraining are used in order to overcome concept drift due to climatic and intervention variability. Besides the accuracy of predictions, the interpretability and operational relevance are a priority of the study because of the usage of explainable AI tools like analysis of feature importance, SHAP values, and LIME visualization. To assess model performance, we measure such metrics as accuracy, F1-score, ROC-AUC, and precision-recall and complemented by ablation studies to evaluate the strength of and contribution of a specific methodological improvement. The hybrid ensemble approach will be based on the ability of the generalization of the Random Forest with the bias-correcting power of the XGBoost to create a transparent/scalable/adaptive ecologically specific system in the highland environments. Besides technical improvement, the research advances the principles of responsible and ethical AI by focusing on interpretability, fairness, and data governance. The expected deliverable is the locally calibrated predictive system, which can be used to support timely, evidence-based public health decision-making in the Kisii County and can be used to offer a transferable methodological framework of malaria surveillance in ecologically similar highland areas across Sub-Saharan Africa.Item An isolation forest model for anomaly detection of Data exfiltration in network traffic Case study – academic institutions network environments(KCA University, 2025) Arusei, Mike K.Academic institutions are increasingly relying on interconnected networked and to manage critical data and services making them vulnerable to cybersecurity threats such as data exfiltration. Traditional security infrastructures often fail to detect these emerging threats, especially within resource constrained academic environments lacking sufficient expertise. To address this, an anomaly detection model was designed using unsupervised Isolation Forest algorithm, which analyzes key network features identify abnormal outbound traffic indicative of data breaches. The model was evaluated on the CICIDS2017 dataset, focusing on real-world infiltration scenarios, with a case study of academic institutions to ensure contextual relevance. Using recall, precision, and F1-score metrics, the model demonstrated effective detection capabilities. Its significance lies in providing a scalable and practical network security solution for academic institutions, supporting compliance with data protection regulations. However, limitations include reliance on the representativeness of the dataset and adaptability to emerging attack patterns. Future work should include exploration of continuous model refinement and integration with broader security frameworks.Item An ensemble learning model for prediction of artificial insemination outcomes in Kenyan dairy cows: a case study.(KCA University, 2025) Rutto, Erick K.Artificial insemination is a vital reproductive technology for smallholder dairy systems, yet its adoption in Kenya and East Africa remains low majorly due to poor success rates influenced by complex management, animal-related, and farmer-related factors. The ability to predict the outcome of each insemination based on the available influencing factors using a decision support tool will boost the adoption of this breeding technology. The background of such a tool is machine learning models. This case study sought to develop an ensemble machine learning model to predict the artificial insemination outcome of dairy cows in smallholder systems. This was addressed through four objectives: assessing influencing factors for artificial insemination outcome, evaluating existing machine learning models using smallholder farmer data, developing a tailored ensemble model for predicting insemination outcome in dairy cows in smallholder systems, and testing performance on heterogeneous data from small scale farmers. The study utilized data on pregnancy diagnosis outcome obtained from smallholder dairy farmers in Kakamega and Kisumu counties of Kenya. Additional data from regions with the same ecological zones in Tanzania and Ethiopia were incorporated to corroborate the findings in the east African context. A total of 1347 pregnancy diagnosis outcome records were used to test the predictive ability of five models such as Support Vector Machines (SVM), Naive Bayes (NB), K-Nearest Neighbors (KNN), Random Forest (RF) and Decision Trees (DT). The outcomes of these models were then fitted to Logistic Regression (LR) stacked ensemble model and its performance compared with XGBoost stacked ensemble. Results from feature importance analysis identified estrus synchronization (0.23), cow age (0.16), body condition score (0.13), membership to cooperative (0.09), and fodder growing (8.81%) as key predictors, while conventional health factors like vaccination and deworming showed minimal impact. Assessing existing base models for this type of data it was found that Random Forest (RF) outperformed all others (accuracy: 0.88). Logistic Regression-based ensemble achieved robust results (accuracy: 0.86) and exceeded XGBoost ensembled model but was outperformed by RF. The study highlights the importance of management decisions over traditional health interventions and the strengths of tree-based models in handling such type of data providing a framework for data-driven reproductive management.Item Predictive model to forecast desert locust outbreaks in Kenya using maximum entropy(KCA University, 2025) Chepkwony, Noah K.Desert locusts (Schistocerca gregaria) are one of the most destructive transboundary pests, posing significant threats to food security, livelihoods, and vegetation. In Kenya, a severe outbreak of desert locust outbreak occurred between December 2019 and June 2021, causing extensive damage to crop and vegetation, specially in the eastern and northeastern parts of the country. Using forecasted climate and environmental data as well as historical occurrence data, it is possible to Predict possibility of an outbreak which can facilitate relevant stakeholders to put in place necessary measures to mitigate the effects. This prediction can help enhance early warning systems by facilitating timely intervention towards mitigating risks efforts. This research study aimed at coming up with a predictive model for desert locust outbreaks in Kenya using the MaxEnt algorithm and historical presence data together with environmental variables such as precipitation, soil moisture, temperature, and vegetation indices to identify areas susceptible to infestations. The research used used Maxent algorithm and latest technologies of GIS and machine learning techniques to generate maps that classify areas in terms of risks levels (low, medium, high) based on climate data and historical locust occurrence data. The output will help enhance locust monitoring and forecasting, providing critical insights for policymakers, stakeholders, and farmers. The output includes a validated prediction model, maps, and recommendations for locust control strategies. The findings revealed that precipitation and soil moisture were the strongest predictors of habitat suitability, followed by temperature and vegetation indices. The MaxEnt model produced validated habitat suitability maps, classifying areas into low, medium, and high-risk zones. High-risk areas were concentrated in northeastern and eastern Kenya, aligning with regions historically affected by locust invasions. These results demonstrate that combining presence-only data with climatic and environmental predictors provides reliable forecasts of potential outbreak zones. The study concludes that the predictive model and generated risk maps can strengthen early warning systems, guide surveillance and control operations, and support policymakers, stakeholders, and farmers in mitigating the impact of desert locust outbreaks.Item Design and evaluation of machine learning - based framework For fraud detection in Kenya’s digital payments ecosystem(KCA University, 2025) Gitau, Denis G.The rapid growth of Kenya’s digital payments ecosystem has revolutionized financial transactions, enabling greater financial inclusion through fintech innovations such as Point-of-Sale (POS) systems, mobile wallets, and e-commerce platforms. However, this expansion has also increased exposure to fraud, as cybercriminals exploit digital vulnerabilities and high transaction volumes to execute increasingly sophisticated schemes. Conventional rule-based fraud detection systems, which rely on static thresholds and predefined patterns, have proven inadequate in addressing evolving fraud tactics. They often result in high false-positive rates and delayed responses, ultimately compromising customer trust and financial integrity. This study presents the design and evaluation of a machine learning based framework for fraud detection tailored to Kenya’s digital payments ecosystem. Using anonymized transaction data from Pesapal Ltd, a leading regional fintech provider, the research applies a range of supervised learning algorithms including Random Forest, Gradient Boosting, Logistic Regression, Naïve Bayes, Decision Trees, KNearest Neighbors, and Neural Networks to identify the most effective approach for real-time fraud detection. The study follows the CRISP-DM (Cross-Industry Standard Process for Data Mining) methodology and incorporates feature engineering, Synthetic Minority Oversampling Technique (SMOTE), cost-sensitive learning, and explainability techniques to enhance model robustness and interpretability. Model performance was evaluated using precision, recall, F1-score, AUC, and costbased metrics to capture both statistical accuracy and business implications. After applying SMOTE, Random Forest demonstrated a superior balance between detection sensitivity and false-positive control, achieving Precision = 52.54%, Recall = 52.77%, F1-score = 52.66%, and PR-AUC = 53.55%, outperforming other models in identifying fraudulent transactions. The SHAP-based explainability analysis further highlighted the dominant role of transaction geography, processing bank, and card type in predicting fraud. The results highlight the effectiveness of ensemble learning techniques for fraud detection in imbalanced financial datasets and demonstrate the value of explainable AI (XAI) in enhancing transparency and regulatory compliance. The study contributes to the ongoing discourse on financial cybersecurity in African fintech ecosystems by offering a scalable, interpretable, and contextually relevant fraud detection framework suited for real-time digital payment environments.Item A negative binomial regression analysis of road traffic Deaths in Nairobi(KCA University, 2025) Sitati, Daphene N.The majority of people in Kenya travel by road, with a sizable portion of the population relying on various modes of transportation for daily commuting. Road Traffic accidents (RTAs), also known as collisions or crashes, happen when a car hits another car, a pedestrian, a road barricade, an animal, or any stationary object like a tree or electricity pole. (RTAs) are becoming common in Kenya and other African countries due to population growth and increased motorization. There are significant barriers to road safety in Kenya, as highlighted by the 2023 economic survey, which revealed an astounding 4,690 fatalities on Kenyan roads in 2022. The proposed study's goal was to analyze Road Traffic Deaths in Nairobi using a Negative Binomial Regression Model. The model was validated using the Pearson Chi Squared Statistics and the root mean square error. Majority of studies that have been conducted have analyzed the effects of weather parameters, road structure, and environment on traffic accidents; however, human factors like gender, age, and drug use are frequently interrelated when determining the death rate of an accident. The data used in the study was secondary between the period of 2017- 2023, derived from Traffic base commanders in Nairobi County, and National Transport Safety Authority (NTSA) accident data. The analysis revealed a decline in road fatalities from 2020 to 2023. The study found that drivers were responsible for the majority of fatal road accidents, with most occurring on Saturdays and Sundays. The cause of death coefficients were as follows: drivers (0.3645), pedestrians (0.4439), pedal cyclists (0.4914), and vehicle defects (0.5871). The study's findings are expected to contribute to enhanced road safety measures, guiding the development of better policies, rules, and interventions aimed at reducing the death toll and its adverse effects on Kenya's economy and society.Item Predicting campus admission through assessment of soft skills using random forest algorithm(KCA University, 2025) Muthui, Dennis M.This study develops a machine learning model to predict college admission success in Kenya by assessing soft skills using the random forest algorithm. The research addresses the growing importance of soft skills in academic and professional success, and current limitations in evaluating these abilities during admissions. The study identifies key soft skills, creates a comprehensive assessment tool, develops and tests a random forest model, and evaluates its performance, interpretability, and fairness. The methodology involves a quantitative predictive modeling design, employing stratified random sampling and rigorous data collection procedures. Results indicate that soft skills, particularly communication and problem-solving, are strong predictors of admission success, often outweighing traditional academic metrics. The random forest model achieved 98.36% accuracy in predicting admissions outcomes, with mathematics performance emerging as the most influential factor (22% importance), followed by GPA (18%), KCSE scores (16%), and science grades (15%), while soft skills showed more modest but meaningful contributions (communication 8%, leadership 5%, problem-solving 4%). The model demonstrated consistent performance across demographic groups, with perfect equal opportunity across gender, school type, location, and school level categories. However, the model reflected existing demographic disparities in admission rates that mirror broader equity challenges in educational access. The study concludes that incorporating soft skills assessments in admissions processes could provide a more holistic evaluation of applicants, though current practices continue to prioritize traditional academic achievement. Recommendations include integrating soft skills development in secondary education curricula and incorporating structured soft skills assessments in university admissions processes. This research contributes to the ongoing dialogue about evolving higher education admissions to better align with 21st-century workforce needs while promoting fairness and transparency in the admission process.Item A paired-algorithm clustering model for describing field staff Deployment in non-governmental organizations (ngos).(KCA University, 2025) Nyakado, Manasses N.This research addresses the inefficiencies and challenges faced by non-governmental organizations (NGOs) in deploying field staff, focusing on the manual processes prevalent in the current systems and leveraging on the possibilities offered by predictive machine learning algorithms. The problem stems from time-consuming and error-prone manual data entry methods, hindering optimal resource allocation. Our objective is to develop and implement a machine learning clustering algorithm to automate the field staff deployment process. By leveraging data analytics – hierarchical and k-means machine learning algorithms – we aim to enhance the efficiency and accuracy of deployment, leading to improved allocation of personnel and resources. The expected outcome is a streamlined deployment system that significantly reduces errors, minimizes time consumption, and maximizes overall operational efficiency in NGO field operations. The project outcomes will also inform advances in the use of combined methods in clustering machine learning algorithms and data analytics.