School of Computing and Mathematics (SCOM)

A hybrid prophet -LSTM model for drought forescasting in Machakos County.

2025-01-01T00:00:00Z

A hybrid prophet -LSTM model for drought forescasting in Machakos County. Katumo, Eric Kioko In Kenya, drought remains a challenge in its arid and semi-arid lands (ASALs) where Machakos County located in the Eastern region, it experiences such events every three to four years. Crop yields, livestock losses, and water shortages are highly negatively affected by these recurrent episodes. Forecasting models, including ARIMA, SARIMA, Numerical Weather Prediction (NWP), ARIMA, and SARIMA are the conventional models used to forecast drought in the region and they highly struggle to capture the complexity of drought patterns resulting to errors of more than 30% due to uneven seasonality and nonlinear climate dynamics. To overcome these limitations, this study created a hybrid forecasting model that combines strengths of both Prophet and Long Short Term Memory (LSTM) models. Daily and Monthly climate data from 2014 to 2024 with variables; humidity, precipitation, temperature and wind speed was used for model training and validation. In this framework, prophet was applied to extract long-term trends and seasonal effects, while LSTM was employed to capture nonlinear residuals variations, improving overall prediction accuracy. After analyzing the findings of the study using drought indices, it was discovered that the LSTM model with SPI-6, produced the most accurate findings with MAE = 0.3847, RMSE = 0.5410, R2 = 0.7833), which means that it explained 78.33% of the variability in drought. The Prophet-LSTM model produced great results, with MAE = 0.4788, RMSE = 0.5765, and R2 = 0.7539. It performed better than baseline Prophet model and kept the forecast stable even in the presence of data gaps. Eight drought events between 2014 and 2024 were discovered with an average duration of 2.5 months and 13-months of drought from October 2022 to October 2023. In general, the hybrid model offers a more reliable tool for forecasting drought in Machakos County and shows potential for application in other ASAL regions. The study introduces a tailored Prophet–LSTM framework for semi-arid climates, showing that seasonal decomposition combined with nonlinear learning reduces forecast errors and improves precision over traditional models. Future improvements may include integrating additional climate variables, expanding datasets, and enabling real-time model updates to strengthen drought preparedness and resource planning. Research project submitted to the department of computer Science and information technology, school of mathematics and Computer science, in partial fulfillment of the requirements for the award of the degree of master of science in Data Science of the co-operative university of Kenya.

A model to determine cyber security human vulnerabilities exposure index for MFIs(A case study of Nairobi county,Kenya

2026-06-29T00:00:00Z

A model to determine cyber security human vulnerabilities exposure index for MFIs(A case study of Nairobi county,Kenya Evaline, Njeri Waweru This study developed a model to determine the Cyber Security Human Vulnerability Exposure Index (CSHVEI) for Microfinance Institutions (MFIs) in Nairobi County, Kenya. While people are a critical component of organizational security, they often represent the most significant vulnerability. An integrative literature review identified key human factor vulnerabilities, which were consolidated into three core variables: human error, negligence, and ignorance. A survey was administered to 132 respondents from 52 MFIs, achieving an 85% response rate (n=112). The collected data was analyzed using Spearman's rank correlation and multiple linear regression. The regression analysis produced a highly significant model (F(3,108) = 341.184, p < .05) that explained 90.2% of the variance in the CSHVEI (Adjusted R² = .902).The resulting formula, CSHVEI = -0.062 + (0.167 × Human Error) + (0.539 × Negligence) +(0.324 × Ignorance), was implemented and validated via a web-based prototype. The study concludes that negligence is the most weighted factor influencing human vulnerability. The model provides MFIs with a tool to quantify their human factor exposure, enabling targeted interventions to strengthen their overall cybersecurity posture. A thesis submitted to the department of computer science and Information technology in the school of computing and Mathematics in partial fulfilment of the requirements for The award of the degree of master of science in cyber Security of the Co-operative University of Kenya.

Regression algorithm-based machine learning model for Apartments’ price prediction in Nairobi city.

2025-01-01T00:00:00Z

Regression algorithm-based machine learning model for Apartments’ price prediction in Nairobi city. Gift, Merqular Odieny. The real estate market of Nairobi has been booming quickly with the price of apartments depending on location, amenities, and the market forces. Conventional approaches to valuation that use the past and market judgement can hardly be accurate or efficient. This paper presents and verifies a machine learning model to estimate the price of apartments in Nairobi. Online sources and Kenya National Bureau of Statistics (KNBS) were used to gather data and three regression algorithms of Linear Regression, Random Forest (RF), and Gradient Boosting Machines (GBM) were compared. The models were trained, tested and validated to find out the predictive accuracy. These findings indicated RF and GBM were more successful than Linear Regression and Support Vector Machine (SVM) with an accuracy of 86.30 and 84.40, respectively. The importance of features analysis allowed determining the apartment size as the key factor that determines the price after which came the number of bedrooms and bathrooms. The research paper suggests that RF and GBM should be used to create a web-based prediction tool, which will provide real estate experts and investors in Nairobi an accurate, transparent, and reliable pricing model. In general, the results prove that machine learning models are effective to predict the non-linear behaviour of apartment prices, and they are better than traditional valuation methods. A project submitted to the department of computer Science & information technology in the school of Computing and mathematics in partial fulfillment of the Requirements for the award of the degree of master of Science in data science of the co-operative university of Kenya.

Multi-task machine learning model with explainability for Personalized nutrition in Kenya.

2025-01-01T00:00:00Z

Multi-task machine learning model with explainability for Personalized nutrition in Kenya. Wamucii, Johnson Kiiru This project created and contrasted between Multi-Task Learning (MTL) and Single-Task Learning (STL) models in order to make personalized recommendations on nutrition based on an explainable AI method. Its overall goal was to design, implement, and test a predictive machine- learning model, which at once forecasts daily calorie consumption, protein, carbohydrates and fats at once based on individual-specific data, and explicitly compares MTL and STL architectures.The research utilized the publicly available Personalized Medical Diet Recommendations Dataset that consists of 5,000 individual records with the variables of demographic, clinical, lifestyle, dietary, and wearable-device. Complete records were kept after cleaning and preprocessing to be used in model training and testing. A multi-task deep-learning model with shared layers was trained parallel to four single-task deep-learning models. Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and coefficient of determination (R2) were used to measure performance.Findings indicated that the MTL model performed better than the STL models when all four nutritional targets were used with the highest difference observed in protein prediction (R2 of 0.96 vs 0.65) and the overall efficiency of the parameters were achieved to about 48 percent. SHAP (SHapley Additive explanations) was incorporated to give an easy to understand feature-attribution explanation that showed BMI, weight, dietary habits, and current nutrient intake as the most significant predictors. According to the study, the performance of a multi-task learning model in individualized prediction combined with explainable AI is more accurate, efficient, and transparent than the use of the traditional single-task models. The framework which has been obtained offers an effective, reliable instrument of offering customized, evidence-based nutrition guidance that can help improve user comprehension and compliance. A project submitted to the department of information Technology and computer science in the school of computing and Mathematics in partial fulfilment of the requirements for the Award of the degree of master of science in data science of the Co-operative university of Kenya.