Faculty of Science, Engineering and Technology (FSET)

Permanent URI for this communityhttps://repository.chuka.ac.ke/handle/123456789/156

Browse

Now showing 1 - 20 of 59

A hybrid deep learning model for intrusion detection in cloud-based implantable medical devices
(Chuka University, 2015) Kirimi James
The rapidly evolving technologies in the healthcare sector, such as implantable medical devices (IMDs), require advanced security solutions that leverage the intelligence capabilities of these technologies while ensuring optimal safety and reliability. The IMD technology redefines healthcare service delivery by offering timely interventions, minimally invasive treatment options, and continuous patient condition monitoring to improve quality of life. Despite these achievements, IMDs face unauthorised access, data manipulation, and denial-of-service attacks, which conventional security solutions are limited in handling due to resource constraints within IMD ecosystems. As a result, different machine learning and deep learning frameworks have been proposed for real‐time threat detection. However, they still suffer from overfitting, slow inference, and excessive resource demands, hindering their effective integration into the IMD ecosystem. The study's primary goal was to design and develop a hybrid of deep autoencoders, convolutional neural networks, and long short-term memory (LSTM) strategies to provide a comprehensive detection model that reduces inference time for deployed models while enhancing performance. Autoencoders provide the fundamental architecture of the detection model, while convolutional neural networks are used in the encoder and decoder for simplicity and to capture nonlinear data effectively. The Long Short-Term Memory captures temporal dependencies in the model, enhancing overall detection capabilities. The study adopted an experimental approach, developing a hybrid deep autoencoder model to test its performance against convolutional neural networks, long short-term memory, and other conventional machine learning techniques. The results demonstrate that the hybrid model outperformed standalone models, achieving high accuracy scores across the datasets. The best model in the ICU dataset achieved 100% accuracy, precision, recall, and F1 score, and a false positive rate of 0.00%. The WUSTL had an accuracy of 79.32%, a recall of 79.92%, a precision of 79.41%, a specificity of 79.24%, and a false positive rate of 20.59%. The Edge IIoT dataset had a recall, F1, and accuracy of 96.87%, a precision of 96.94%, a specificity of 96.88%, and a false-positive rate of 3.12%. The model’s inference time was substantially reduced compared to the standard deep autoencoder model across the datasets, providing a lightweight detection environment for the intrusion detection system.
A hybrid of deep auto-encoder and feature embedding model for an e-commerce recommender system
(Chuka University, 2024) Ireri Justin Murithi
Recommender systems aim to predict user interests and suggest products that are likely to be of interest. These systems are widely used across various platforms, including online shopping, streaming services, and music stores, to provide personalized suggestions. Traditional machine learning-based models, such as collaborative filtering and content-based algorithms, often face challenges like low accuracy, data sparsity, and the cold start problem. The cold start problem occurs when a system lacks sufficient data to make accurate recommendations for new users or items. This study specifically focuses on addressing the visitor cold start problem, where the system does not have prior information about the new user’s preferences or behavior, making personalized recommendations difficult. To address this issue, a model was developed using deep auto-encoders integrated with feature embedding (DAE-FE), designed to improve item prediction accuracy for new users in an e-commerce recommender system. The model introduces an embedding layer after the dropout layer in the deep neural network, which automatically captures user data points such as time and location. These data points help in constructing a user profile necessary for prediction. This feature not only improves the accuracy of item predictions but also speeds up the process by filling in missing data for new users, allowing the system to proceed directly to prediction. An experimental research design was employed to compare the performance of the developed model with previous models that relied solely on provided datasets. In the experiment, user location and time of login were used as independent variables, while model accuracy served as the dependent variable. The model was trained and tested using the MovieLens 100k dataset, which was adapted to meet the requirements of the DAE-FE model. The hybrid model achieved a mean squared error of 0.0241 and a root mean squared error of 0.1443, indicating minimal deviation from the actual values. As a result, the model attained approximately 96% accuracy in predicting recommendations for cold start users. Overall, the model demonstrated strong performance and appears to be a promising solution for the cold start problem in ecommerce systems. The research found that incorporating more side information from users and items on the dataset during the model's training will yield more accuracy in item prediction.
A hybrid of residual network and inception neural network model for wildlife detection and identification
(Chuka University, 2025) Malach Obisa Amonga
Machine learning has significantly transformed various domains, with deep learning architectures playing a crucial role in computer vision applications. Convolutional neural networks (CNNs) have demonstrated remarkable success in image classification and object recognition tasks. However, traditional CNN architectures often exhibited limitations in handling complex feature extraction and generalization, particularly in wildlife identification where intra-class variations were high. The challenge in wildlife identification arose due to factors such as varying lighting conditions, occlusions, background clutter, and pose variations, which made it difficult for sinle model architectures to achieve high accuracy and robustness. This study sought to address these challenges by first designing and implementing individual Residual Network (ResNet) and Inception models to establish baseline performance, and then developing a hybrid ResNet-Inception model aimed at enhancing feature extraction, optimizing classification performance, and improving generalization capabilities in wildlife identification tasks. The Animals with Attributes 2 (AwA2) dataset was used to train and evaluate the models, and their performance was assessed using standard classification metrics, including accuracy, precision, recall, and F1-score. The WildlifeReID-10k dataset served as an external validation set. The hybrid approach leveraged ResNet’s ability to mitigate vanishing gradient problems through residual learning and Inception’s capability to capture multi-scale spatial features, thereby creating a more robust and efficient architecture. The results demonstrated that ResNet101 achieved an accuracy of 93.5%, Inception v3 achieved 95.6%, while the proposed hybrid model achieved 98%, confirming its superior performance in distinguishing visually similar species and enhancing generalization. The findings of this study provide a practical contribution to biodiversity conservation by enabling improved automated wildlife identification systems that support ecological monitoring, species recognition, and anti-poaching surveillance. By addressing the limitations of singlemodel approaches and demonstrating the advantages of hybrid deep learning architectures, the study sets a new benchmark in wildlife identification and reinforces the integration of artificial intelligence into environmental conservation practices.
AB INITIO STUDY OF STRUCTURAL AND PIEZOELECTRIC PROPERTIES OF HAFNIUM DOPED BISMUTH SODIUM POTASSIUM TITANATE
(Chuka University, 2023-10) MWANZIA BONFACE MUTUKU
Piezoelectric materials have gained increased attention in the recent times due to their significant technological applications. These materials are widely used to make ultrasound transducers, sensors, actuators and others are used for energy harvesting. Due to its brilliant piezoelectric properties, Lead Zirconate Titanate (PZT) is mostly used with a piezoelectric constant of 𝑑33 = 374 𝑝𝐶/𝑁 from experimental reports and 306 − 314 𝑝𝐶/𝑁 from theoretical studies. However, due to the toxic nature of lead oxide which is formed when PZT is being manufactured, there is increased effort in development of lead-free materials. Several classes of materials have recently been studied and are now being considered as potential alternatives to PZT. Lead free perovskite systems such as Bismuth Sodium Potassium Titanate (BNKT) have been developed, with a piezoelectric constant 𝑑33 = 157 𝑝𝐶/𝑁 . However, the main drawback of this system is that it is highly corrosive and has a low piezoelectric constant compared to PZT. In the quest to provide suitable alternatives, dopants such as zirconium have been used, which improved the piezoelectric constant of BNKT up to203 𝑝𝐶/𝑁. Hf which possesses similar physico-chemical properties as zirconium has led to an improvement in the piezo electric constant of other piezoelectric systems such as in hafnium doped Barium Titanate (BT). It has an added advantage of being extremely resistant to corrosion, which is expected to mitigate the corrosive nature of BNKT. In this study, hafnium has been incorporated in BNKT so as to engineer an alternative material suitable for piezoelectric applications. Density Functional Theory (DFT) method was used to predict the structural and piezoelectric properties of hafnium doped BNKT, starting with those of Bismuth Sodium Titanate (BNT) and BNKT. The exchange and correlation was taken as the Generalized Gradient Approximation (GGA). The optimal lattice parameters for BNT were found to be 𝑎 = 5.57 Å and 𝑐/𝑎 ratio of 2.50 for the conventional cell, having space group R3c space group number 161. Piezoelectric constant for this system was found to be 97.67 pC/N. This structure was adopted for doping and further calculations. Potassium doped bismuth sodium titanate was modelled using VESTA software and its optimized lattice parameter was found to be 𝑎 = 5.60 Å. Piezoelectric constant for this system was found to be 147.42 pC/N. Hafnium doped BNKT had an improved piezoelectric constant of 205.52 pC/N for 3% hafnium doping, which decreased to 163.22 pC/N at the level of 6% doping. The results shows that small amounts of hafnium improved the piezoelectric constant of BNKT from 147.42 pC/N to 205.52 pC/N. Elastic and elastic compliance full tensors for these systems was also generated with elastic constants of C33 = 286.48 Gpa, 282.13 Gpa, 257.193 Gpa and 276.43 Gpa for BNT, BNKT, 3% Hf doped BNKT and 6% Hf doped BNKT respectively. This study concludes that doping BNKT with hafnium indeed improves the piezoelectric properties of BNKT. This makes this material more useful in energy generation since high piezoelectric constant leads to efficient mechanical – electrical energy conversion in the piezoelectric materials.
ACSC 479: TECHNO-PRENEURSHIP AND CAREER CHOICES
(Chuka University, 2023-04-19) Chuka University
An efficient detection model of zero-day web application attacks based on convolution neural networks and deep auto encoders
(Chuka University, 2024) Tuei Kevin Kirui
The need for secure and trustworthy information systems has taken center stage and proven critical in supporting teleworking, online teaching, and research services. Artificial Intelligence (AI) is the primary driver of the 6th generation of computing, and innovations with applications of AI in computer vision, gaming, robotics, and security. Zero-day web application attacks take advantage of web application software weakness for as long as the developer is unaware and has not developed a mechanism to eliminate the weakness. Zero-day attacks leave vulnerable users grappling with data loss and have the propensity to push an organization out of business. Current zero-day attack detection methods built on signature-based or anomaly-based methods are inefficient in combating these attacks since they rely on previously detected weaknesses for signatures and a deviation from normal behavior for anomaly detection. These methods result in detection rates below 80%, meaning the propensity of Zero-day attacks going undetected is 20% or lower. The application of machine learning techniques has proven to be efficient because these techniques can continuously learn from the code as well as its execution to detect security breaches and trigger an alarm. With the need to improve these techniques, a novel classification model needs to be developed to increase the detection rate further and reduce the false alarm rate. This study applied a hybrid of two machine learning methods, Convolution Neural Networks and deep autoencoders, to develop a classification model that significantly increases the detection rate of zero-day attacks. The KDD'99 Dataset is a comprehensive repository of fully labeled intrusion detection records that was used to develop, test and validate the model. This dataset simulated real-world scenarios and assessed the model's performance under different intrusion scenarios. The Average Detection Rate, Accuracy and F1 score metrics were used to evaluate the model. The hybrid CNN-Deep Autoencoder model had a detection rate of 0.895 against 0.887 of the Fully Connected Network (FCN) with sampling and 0.885 of the pure CNN model. The accuracy and F1-score of the hybrid CNN-Deep Autoencoder were 0.973 and 0.971 respectively. The Hybrid Model of CNN and Deep Autoencoder is efficient in detecting Zero-Day Attacks making it possible for Software Developers to patch their systems sooner resulting in minimal dwell time.
An enhanced convolutional neural network model for translating Kenyan sign language into text in english
(Chuka University, 2024) Muthui Nancy Njoki
Most people communicate effectively and socialize through verbal means, such as talking. However, mute and deaf people cannot interact with society through speech. So, they use the non-verbal modes of communication. Non-verbal communication is a sort of usual body movements, hand gestures, and facial expressions like sign language, and this needs translation according to the specific patterns that the gestures and facial expressions or positioning of the hands, fingers, and arms carry with them during sign language. While it bridges a gap between those who can hear and those who cannot, it is by no means universally comprehended, thus standing as a barrier that leads to frustration and social exclusion of deaf people. As such, a translation tool may help convert sign language into easily understandable written language that will facilitate smooth communication between hearing and hard-of-hearing persons. While lots of research is going on in the area, little attention has been given to translating Kenyan Sign Language into some of the commonly spoken languages in Kenya. Besides, most translation tools face several challenges due to changing environmental conditions and the movement of a person while performing sign language, leading to changes in background lighting. This work translates KSL into English text through the experimental approach using a deep learning CNN model, DenseNet121, preprocessed by Contrast-Limited Adaptive Histogram Equalization. This architecture has been developed, trained, and tested on the dataset provided by the Kenyan Sign Language Classification Hackathon with an accuracy of 91.5%. The proposed model will bridge communication gaps and help include people who are hard of hearing in educational, health, and employment opportunities.
ANALYSIS OF PRODUCTION, SOCIO-ECONOMIC AND INSTITUTIONAL FACTORS AFFECTING TECHNICAL EFFICIENCY OF SMALLHOLDER BANANA PRODUCERS IN KIRINYAGA CENTRAL SUB-COUNTY, KENYA
(Chuka University, 2023-10) MATIVA JACKSON MKENYE
Banana provides food, nutrition security and income for most households and is fourth most popular food crop in the world after wheat, maize and rice. Despite its significance, full potential of banana production in Kenya remains unexploited by smallholder producers. This is as a result of low technical efficiency especially in utilization of farming inputs and producer specific factors like production, socio-economic and institutional factors among others. In Kirinyaga County, the actual banana production is at 4-18 tonnes per acre against the potential of 30-40 tonnes. Due to the limited supply of resources for production, attainment of highest possible levels of technical efficiency is key to achieving sufficiency in banana farming. This study aimed at analyzing the effects of production, socio-economic and institutional factors on technical efficiency of smallholder banana producers in Kirinyaga Central Sub-County, Kenya. The study used a cross-sectional research design and targeted a population of 24,440 smallholder banana producers. Multistage sampling technique was employed where purposive sampling and simple random sampling methods were used in some stages to sample respondents in the study area. A sample of 402 smallholder banana producers were selected. Using a questionnaire, primary data on production, socio-economic and institutional factors affecting technical efficiency of banana production was collected. The data was then analyzed using Stata version 17 and SPSS version 25. Descriptive statistics were used to describe the production, socio-economic and institutional factors of the smallholder banana producers. A stochastic frontier analysis approach was used to model the technical efficiency level using the Cobb-Douglas function. The stochastic production function of the Cobb Douglas function was estimated using the maximum likelihood estimation technique. The study showed that the level of banana production technical efficiency among the smallholder producers varied between 0.9% to 95.5% and average technical efficiency of 83.1%. According to the model parameters calculated, land set aside for production of banana, banana suckers and agrochemicals were significant production factors in banana cultivation at 5% significance level. The study found that agrochemicals and planting materials had positive effects on technical efficiency whereas land size had a negative impact on technical efficiency. This implied that increasing the amount of land set aside for banana production by an acre reduces the amount of banana harvested by 0.438 kgs while increasing the amount of planting materials and agrochemicals used by one unit increases banana production by 1.315 and 0.155 kgs, respectively. The study found that decision makers’ age and size of the household had negative effects on banana production technical efficiency whereas education, experience, producer group membership and market access had positive effects. The study recommends people with high levels of education to venture into banana production. In addition, producer group formation and membership be encouraged for the benefit of increasing technical efficiencies. The inefficient producers are advised to increase their present output by enhancing technical efficiency as a result of the inefficient utilization of their resources throughout production.
APPLICATION OF QUEUING THEORY FOR OPTIMAL CUSTOMER CENTRICITY TO THE BANKING SECTOR IN KENYA
(Chuka University, 2023-10) JUMA SAMWEL KISIANG’ANI
Long queues and waiting times are common in banks, resulting in customer dissatisfaction and low customer retention. The study applied a descriptive research design to investigate queuing dynamics in a banking hall at a commercial bank in Kenya. A single server system (M/M/1) queuing model was used to estimate the average waiting time, system intensity, service time, and optimal number of staff during peak and off-peak periods (July). The study used secondary data on daily waiting times, service times, the number of customers, and servers for May and July 2019, 2020, and 2021 during working hours between 8.30 a.m. and 4 p.m. on Monday to Friday and 8:30 a.m. and 12 p.m. on Saturdays. Data analysis was done using R and Excel. The research findings indicated that the peak periods (May) recorded an average waiting time (AWT) of 13 minutes, 35 seconds in 2019, 10 minutes, 14 seconds in 2020, and 8 minutes, 36 seconds in May 2021. In the off-peak periods (July), an AWT of 3 minutes, 46 seconds, was registered in 2019, 5 minutes, 12 seconds in 2020, and 7 minutes, 42 seconds in 2021. An average service time (AST) of 1 minute 52 seconds in May 2019, 2 minutes 34 seconds in May 2020, and 2 minutes 27 seconds in May 2021. In the off-peak periods (July), an AST of 3 11 seconds was registered in 2019, 3 4 seconds in July 2020, and 2 43 seconds in July 2021. Overall, the system intensities are low to moderate, with the COVID-19 pandemic severely impacting the peak period more than the off-peak. In the peak periods, the service rates averaged 33, 24, and 25 persons per hour in May 2019, May 2020, and May 2021. The respective system intensities were 0.534, 0.360, and 0.492. In the off-peak periods, the average service rates were 19, 20, and 23 persons per hour in July 2029, July 2020, and July 2021. The respective associated system intensities of 0.535, 0.461, and 0.487. From the pooled data for 2019 and 2021, the study recommends that banks operate with an AWT of 6 minutes, 24 seconds, and an AST of 3 minutes. Further, the study established that a bank could work with an optimal four servers with an AST of 2 minutes, 35 seconds (a service rate of 20 people per hour), and achieve a moderate average service intensity of 0.552.
Assessment of physico-chemical properties, bacterial contamination, and water quality interrelationships in lanyiru spring, meru county, kenya
(Chuka University, 2025) Ikiao Paul Mwongela
Reliable access to clean water is critical for public health, economic growth, and environmental sustainability. However, rural communities in Kenya continue to face water quality challenges stemming from contamination, inadequate monitoring, and fluoride pollution. Springs like Lanyiru in Meru County are commonly assumed safe based on appearance, yet the absence of regular assessments poses unrecognized health risks. This study assessed the physicochemical and microbial quality of Lanyiru Spring water to determine its suitability for domestic use. Samples were collected from four sites along Lanyiru Springs at the middle of wet and dry season (September 2025) and were analyzed through both in situ measurements and laboratory procedures. Physical parameters including pH, temperature, turbidity, conductivity, and total dissolved solids were measured on-site with calibrated instruments to ensure accuracy. Chemical analyses, encompassing cation and anion quantification, were conducted in the laboratory following APHA and EPA standard methods, employing techniques such as Flame Atomic Absorption Spectrometry (FAAS), Inductively Coupled Plasma–Mass Spectrometry (ICP-MS), ion chromatography, and ion-selective electrodes. Microbial contamination was assessed via the Most Probable Number (MPN) method for coliform bacteria. Data collected were analyzed using R software (version 4.3.0), with significance determined at α = 0.05, to detect spatial variations and assess water quality relative to health standards. Significant spatial variations (p < 0.05) were observed in key cations, including Pb2+, Fe2+, Cd2+, Cu2+, Ca2+, Mg2+, and K+. Notably, Pb2+ peaked in Sample LP1A and LP1B (0.008 mg/L), Fe2+ at LP4B (10.03 mg/L), and Cd2+ at LP3A (0.041 mg/L), exceeding recommended limits and indicating contamination risks. Cu2+ was highest at LP4B (0.342 mg/L), with elevated Ca2+ and K+ also recorded at LP4A. Anion analysis revealed significant differences in SO4², Cl, and F concentrations across sites. Sample LP4A exhibited the highest SO4² (132.85 mg/L) and Cl (311.90 mg/L), while LP4B showed the highest fluoride level (8.40 mg/L), exceeding WHO guidelines. The Na+ levels showed no significant spatial variation (p > 0.05). Physical parameters showed no significant variation in pH (mean = 10.16), indicating overall alkalinity, but temperature varied significantly, with LP2 recording the highest at 26.87°C. Lanyiru Spring water is contaminated with heavy metals, elevated fluoride, turbidity, and microbial hazards from both natural and human activities, making it unsafe for direct consumption. To protect community health, regular water quality monitoring, pollution control, improved sanitation, household water treatment, and investment in alternative safe water sources are urgently needed.
ASSESSMENT OF QUALITY OF SHALLOW WELLS WATER IN CHUKA SUB-COUNTY, KENYA
(Chuka University, 2020-12) Rugendo, Edith Mwende
Groundwater from shallow wells is an important source of water for domestic and agricultural uses in Chuka Sub-County. However, most shallow wells in Chuka Sub-County are undocumented because approval is not required to sink a shallow well. Groundwater from these shallow wells is therefore susceptible to contamination by both geogenic and anthropogenic sources. This study was conducted to determine the quality of shallow wells water in Chuka Sub-County. Samples were collected from twenty shallow wells in Chuka Sub-County during the dry and wet seasons. The temperature, pH and conductivity were determined in situ using a conductivity meter. The concentration of cations in the water samples was determined using an Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) system. The concentration of chloride and nitrate ions in the water samples were determined using the argentometric and the spectrophotometric method, respectively. The concentration of sulphate ions was determined using a turbidimetric method while the concentration of phosphate ions was determined using a colorimetric method. The data obtained were subjected to ANOVA to test the significance differences using R-3.5.2. Mean comparison was achieved through LSD. The temperatures of the waters were significantly higher during the dry season. The pH of water in most shallow wells during the wet season was within the limits set by KEBS and WHO. However, shallow well water at Karandini (T15) was highly acidic (pH of 3.95) during the wet season. During the dry season, the shallow well waters were slightly acidic (4.42 - 6.44) except for the shallow wells at Ndagani market (T11), whose water was alkaline (pH = 8.75). The shallow wells water at site T11 had remarkably higher conductivities than the other shallow wells in the study area during both the wet (1150 μs/cm) and the dry (1208 μs/cm) seasons. The concentrations of macro-cations (Ca2+ and Mg2+) were higher during the dry season. The concentrations of the trace and toxic cations differed significantly across sites and seasons but were within the limits set by KEBS and WHO. The concentrations of anions (NO3-, SO42-, Cl- and PO43-) were within the limits set by KEBS and WHO for portable water. Groundwater from most shallow wells in Chuka Sub-County is generally safe for domestic uses but regular monitoring for quality is recommended because, the concentration of nitrate in several wells during the dry season was within the range that causes chronic health effects including colorectal, ovarian, thyroid, kidney and bladder cancers.
Bioprospecting and characterization of streptomyces species associated with termites in Kiang’odu section of mount Kenya forest for antimicrobial potential
(Chuka University, 2025) Mugira Tonny Kiriinya
Antimicrobial resistance poses a major challenge to public health, necessitating the search for novel bioactive compounds from unexplored sources. This study aimed to characterize Streptomyces species from the gut of termites inhabiting mounds in a section of Mt. Kenya forest and evaluate their antibacterial activity against multidrugresistant pathogens: Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter cloacae- ESKAPE. Twenty-five pure Streptomyces isolates were identified based on morphological and biochemical traits, including Gram positivity and spore type. Primary screening was done using the cross-streak method, and secondary screening using the agar well diffusion technique. Data on primary screening was analysed using ANOVA to assess differences among termites from various tree trunks. Secondary screening data, expressed as zones of inhibition (mm), underwent ANOVA with Bonferroni adjustments to compare isolates and pathogens. GC-MS–identified compound abundances per Streptomyces isolate were organized into a matrix (isolates × compounds), averaging duplicate entries. Hierarchical clustering using Euclidean distances and complete linkage quantified isolate similarities, visualized via dendrograms. All statistical and clustering analyses were conducted in R Studio version 4.2.2 at α = 0.05 using relevant packages. Results showed strongest antibacterial activity against S. aureus (e.g., isolates MTG-04 and MTG-09 with 14.00 mm inhibition zones, p < 0.0001), moderate activity against K. pneumoniae (MTG-11, 11.17 mm) and E. cloacae (MTG-04, 11.33 mm), and lower yet notable effects against A. baumannii (up to 9.33 mm) and P. aeruginosa (up to 10.83 mm). Secondary screening of crude extracts confirmed these trends, with isolate MTM-18 producing the largest zone against E. cloacae (15.00 mm, p < 0.05) and strong inhibition against K. pneumoniae (14.00 mm). GC-MS profiling identified a diverse range of bioactive metabolites, including fatty acid amides, triterpenoids, and esters, linked to antimicrobial and antioxidant activities. Streptomyces from termite guts in Kiang’ondu section of Mt. Kenya forest demonstrate promising antibacterial activity primarily against Grampositive bacteria, with select isolates (MTG-04, MTM18) identified as prime candidates for further characterization to optimize bioactive compound production and therapeutic potential.
CHARACTERIZATION AND SCREENING OF ANTIBACTERIAL PROPERTIES OF Actinomycetes FROM RIVER TANA AND LAKE ELEMENTAITA, KENYA
(Chuka University, 2023-04) SHIKUKU BONFACE OLOO
The increased prevalence of bacterial infections has been a major challenge to human with devastating high mortality and morbidity rates. This situation has been worsened by increasing antibiotic resistant strains of pathogenic bacteria, reduced effectiveness of antibiotics in the market, and the emergence of new bacterial infections. This study aimed at identification of antibacterial Actinomycetes species using biochemical and molecular methods, screening for their antibacterial secondary metabolite and determination of effect of pH, fructose, sucrose, urea and sodium nitrate on their antibacterial activities. The experiments for this study was laid out in Complete Randomized Design and replicated thrice to determine the difference between the inhibition zones (mm) of isolates against the tests organisms and effects of different levels of pH, sucrose and fructose on antbacterial properties of isolates. The resultant data ( zones of inhibition in millmetres) was analysed using One Way Analysis of Variance and Kruskal Wallis test in SAS version 9.4. A total of six antibiotic producing Actinomycetes species were isolated from river Tana and lake Elementaita and identified through morphological, biochemical and molecular methods.There was a significant (p<0.05) different antibacterial activity of Actinomycetes isolates against Staphylococcus aureus, Salmonella typhi and Escherichia coli. The thin layer chromatography profiling for secondary metabolites in extracts revealed a total of 13 different spots with each having a unique retardation factor. The GC-MS analysis of the extracts revealed 140 different metabolites which have been documented to have antibacterial properties from the six Actinomycetes isolates. There was a significant (p<0.05) effects of different levels of pH and concentration of fructose, urea and sodium nitrate on the antibacterial activity of Actinomycetes isolates against Escherichia coli. The study has revealed different secondary metabolites in unique combinations across the six Actinomycetes isolates with antibacterial activities against Staphylococcus aureus, Salmonella typhi and Escherichia coli. The findings of this study can help in developing new or alternative antibiotics that can be used for treatment of pathogenic and resistant bacteria.
Characterization of clay samples from Murang’a, Nyeri, Embu and Tharaka Nithi Counties for adsorption of cadmium for water purification
(Chuka University, 2025) Thuo Maryrose Wandia
Heavy metal pollution, particularly cadmium (Cd² ), remains a critical environmental challenge threatening water quality, aquatic ecosystems, and human health. Cadmium is widely introduced into aquatic systems through industrial effluents, agricultural runoff, and urban discharges. This study investigated the adsorption potential of locally available clay minerals as cost-effective and sustainable remediation materials. Clay samples were collected from Gakoigo (S), Mukurwe-ini (2A) Gakindu K1, Karurina (K2) and Mbogoni (M). The clay samples were characterized using Atomic Absorption Spectroscopy (AAS), X-ray Diffraction (XRD), and Fourier Transform Infrared Spectroscopy (FTIR). The results from elemental analysis and ANOVA revealed difference in concentrations of iron, magnesium, sodium, and aluminum across the counties, explaining the calculated F-statistics 2.83 (p = 0.048), 3.01 (p = 0.035), 4.51 (p = 0.020), and 5.22 (p = 0.011), respectively. These constituents are vital as they assist, particularly, the samples from Mbogoni (M) and Gakoigo (S), in boosting cadmium (Cd²) adsorption. On the contrary, calcium and potassium were seen to have less influence, evidenced by their F-statistics of 2.45 (p = 0.068) and 2.13 (p = 0.080), respectively. Physicochemical water quality analysis was done from 12 rivers sites within Nairobi County, Kenya. Turbidity ranged between 2.75–95.67 NTU, exceeding WHO’s 5 NTU guideline in urban rivers due to runoff and effluent discharges. Electrical conductivity (556–1123 µS/cm) surpassed WHO limits across all sites, confirming high ionic loading, while dissolved oxygen (0.51–3.06 mg/L) was critically low, pointing to severe organic pollution. Total suspended solids (12–247 mg/L) and TDS (362–736 mg/L) were elevated in urbanized sites, further degrading aquatic health. Cadmium concentrations (0.0105–0.0498 mg/L) consistently exceeded WHO (0.003 mg/L) and KEBS (0.01 mg/L) standards, with highest levels in industrially impacted rivers, highlighting risks of bioaccumulation and human exposure. Batch adsorption studies demonstrated that pH was a key determinant, with maximum efficiency at neutrality (pH 7, 99.81% removal by K1), while acidic conditions reduced removal due to proton competition. Contact time experiments showed rapid uptake within 20 minutes, with equilibrium achieved at 40 minutes (69.16% removal by K1). Adsorbent dosage showed optimum performance at 0.1 g (99.72% removal at 8 ppm), though higher dosages reduced efficiency due to particle aggregation. Desorption confirmed strong Cd binding, indicating chemisorption via ion exchange and surface complexation. Agitation improved uptake by minimizing mass transfer resistance, with peak efficiency (92.29% by K1) at 400 rpm. Temperature exerted a negative effect, with maximum adsorption at 25 °C (97.52% by K1), confirming exothermic behavior. Isotherm modeling revealed Langmuir’s model (Qmax = 2.06 mg/g, KL = 5.44 L/mg, R² = 0.77) better fit the data compared to Freundlich (R² = 0.67), suggesting monolayer adsorption on homogeneous sites. Kinetic modeling indicated pseudo-second order (R² = 0.8496) best described the process, implying chemisorption. Thermodynamic evaluation showed that cadmium adsorption on the clay adsorbent was endothermic, with a positive enthalpy change (ΔH = 17,936.42 J/mol) and a positive entropy change (ΔS = 64.18 J/mol·K). Gibbs free energy (ΔG) remained negative across all temperatures (ΔG = 1.21 to -1.27 kJ/mol), confirming spontaneous adsorption at higher temperatures.
Comparative analysis of ridge logistic regression, artificial neural networks and extreme gradient boosting for predicting loan default rate in Kenya
(Chuka University, 2025) Lemasulani Mary Mpaine
Predicting loan defaults is vital for financial institutions to mitigate losses from non repayment loans. Despite access to extensive borrower data, banks often struggle to forecast defaults accurately due to limitations in traditional parametric models, which assume fixed relationships between predictors and outcomes. These models may fail under changing borrower behavior and economic conditions. This study addresses the need for adaptive models that minimize classification error while controlling complexity. The objective was to compare the predictive performance of Ridge Logistic Regression (RLR), Artificial Neural Networks (ANN), and Extreme Gradient Boosting (XGBoost) in forecasting loan default within Kenya Commercial Banks. A retrospective predictive modeling design was employed using secondary data from the Central Bank of Kenya, covering 2012–2022. The dataset included borrower demographics, loan terms, and repayment status. Analysis was conducted using R -4.5.1 and Python, with model performance evaluated via confusion matrix metrics, Receiver Operating Characteristics (ROC), and Area Under the Curve (AUC). Descriptive analysis revealed class imbalance, variable skewness, and distinct feature distributions, highlighting the need for robust models. RLR identified key predictors such as credit type, loan purpose, gender, and age as increasing default risk, while income and marital status reduced it. However, RLR achieved only moderate accuracy (75.56%), high specificity (99.69%), and very low sensitivity (1.09%), with an AUC of 0.64. In contrast, ANN demonstrated exceptional performance with 99.99% accuracy, perfect AUC of 1.00, and minimal overfitting. The confusion matrix showed 33,686 true negatives and 10,912 true positives, with only three misclassifications. XGBoost achieved 99.9955% accuracy, 100% specificity, and 99.98% sensitivity, with zero false positives and only two false negatives. Its final log-loss of 0.0001736 indicated near-perfect probability calibration. Comparative evaluation revealed that ANN and XGBoost significantly outperformed RLR across all metrics, especially under class imbalance conditions. These findings underscore the superiority of advanced machine learning models in loan default prediction and their potential to enhance risk assessment in Kenyan commercial banks. It is recommended that financial institutions adopt models like ANN and XGBoost to improve predictive accuracy and support data-driven credit decision-making.
A CONVOLUTIONAL NEURAL NETWORKS AND SUPPORT VECTOR MACHINES HYBRID MODEL FOR NUMBER PLATE RECOGNITION
(Chuka University, 2022-03) Kibaara, Peter
ABSTRACT Automatic Number Plate Recognition (ANPR) systems are applied in many fields such as automatic electronic toll collection, car park management and access control, logistics and vehicle tracking, traffic law enforcement and crime resolution amongst others. Motion blur, plate orientation, lighting changes and image noise severely lower the detection speed and recognition accuracy of these systems. The incorporation of machine learning algorithms in ANPRs has seen Convolutional Neural Network (CNN) being used to develop ANPR models with improved performance in license plate detection. CNNs are best suited for image data where the number of features is large such as license plate detection. This is attributed to their design architecture which enables them to perform feature extraction automatically. However, their speed of execution is slow as the model has to learn a lot of features. Support Vector Machine (SVM) is a supervised machine learning algorithm suitable for classification and regression problems with datasets that have a small number of features. It doesn’t scale up well for large datasets with many features. It has demonstrated high speed and accuracy when used for classification in small datasets such as character recognition. The final stage in ANPR is a character recognition phase and involves few features. These two algorithms have been deployed independently, however the concept of combining the two algorithms for ANPR models remains highly unexplored. The research therefore combines the two models (CNN and SVM) to come up with an efficient hybrid ANPR system with improved number plate recognition accuracy. The two models were developed using a deep cascade framework; a CNN with a SoftMax classifier and a hybrid CNN with a SVM classifier. The Universidade Federal do Paraná (UFPR-ALPR) dataset was used to train validate and test the models. Recognition accuracy, precision, recall and F1 score metrics were used to evaluate the model. The hybrid CNN-SVM model had a recognition accuracy of 91.25% against 89.07 % from the pure CNN model. The weighted average precision, recall, and F1-score of the hybrid CNN-SVM was 92%, 91% and 91% respectively, which was better compared to that of pure CNN. The hybrid model was tested for external validity using the Smart Sense Laboratory (SSIG) dataset. The hybrid CNN-SVM model had a recognition accuracy of 91% against 89 % from the pure CNN model. The weighted average precision, recall, and F1 score of the hybrid CNN-SVM was 91%, 91% and 91% respectively which was better compared to that of pure CNN, which had 90%, 89% and 89% respectively.
Determination and removal of selected heavy metals in treated wastewater from Ruai sewage treatment plant for possible agricultural applications
(Chuka University, 2024) Kakuta Peace Kaviti
Fresh water has become scarce and many arid and semi-arid regions in the world suffers from water shortage. Wastewater reuse remains the only reliable and potential source of water. One of the major challenges in recycling of wastewater is the presence of toxic heavy metals which are persistent and non-biodegradable and are known to affect human health. This study sought to synthesize and characterize soda lime and borosilicate waste glass and a composite of Multi-walled Carbon Nanotubes (MWCNTs)/soda lime waste glass adsorbents and utilize them for removal of Pb2+ from wastewater. Glass wastes were collected within Chuka University and MWCNTs purchased from Hongwu International Group Ltd. The adsorbents were washed, dried, functionalized with nitric-sulfuric acid mixture and characterized using FTIR. Wastewater samples were collected in Ruai wastewater treatment plant, Kenya using grab method, transferred to 250 ml plastic bottles and were transported to Chuka university for analysis in a cooler box at 4oC. Standard laboratory procedures of determining the physicochemical parameters were employed. Batch adsorption experiments were conducted to study the effect of contact time, pH, temperature, shaking speed, initial adsorbate concentration and adsorbents dosage on removal of lead (II) ions. Residual Pb2+ concentration was determined using AAS. The findings were: pH 5.5-7.9, Temperature 22.70C-26.10C, Conductivity 526.7- 1209.7µS, turbidity 73- 1000 NTU, nitrates 6.66-25.1 mg/l, Phosphorus 1.16-10.30 mg/l, BOD5 ranged from 10-480 mg/L, COD 90-980 mg/L, TSS 14-422 mg/L, and TDS 244-967 mg/L. pH, temperature, NO3-, BOD5(wet season) results met the WHO and NEMA standards for wastewater reuse in irrigation while EC, turbidity, P, COD, BOD5(dry season), TDS and TSS did not. The results of heavy metals were Ni 0.02-0.22 mg/L, Zn 0.03-1.67 mg/L, Cu 0.01-0.23 mg/L, Cd 0.01-0.05 mg/L, Fe 0.05-7.24 mg/L, Mn 0.14-2.26 mg/L, and Pb 0.04-0.78 mg/L. The levels of Zn, Cu, and Ni were within WHO and NEMA standards while Cd, Mn, Pb and Fe did not meet the threshold at some sampling points. All the metals studied met the FAO guidelines for reuse of wastewater in irrigation. Characterization of the adsorbents was done using FTIR which displayed the dominant functional groups to be silanols, hydroxyls, carboxylic and carbonyl groups. Adsorption of lead (II) ions was conducted using a composite of soda lime waste glass and multiwalled carbon nanotubes, borosilicate and soda lime waste glass. The composite and borosilicate adsorbents reported 100% adsorption of lead (II) ions while soda lime was average. Adsorption of Pb (II) ions followed Freundlich isotherms for borosilicate and soda lime adsorbents with r2 of 0.8665 and 0.9257 while Composite had a better fit in Langmuir isotherm with r2 of 0.9446. Cd (II) and Ni (II) ions did not interfere with adsorption of lead (II), but a stiff competition for the adsorption sites was observed for the case of Mn (II) ions. Regeneration efficiencies of 99.61%, 97.45%, and 99.82% were observed for borosilicate, soda lime, and composite adsorbents. The findings of this study clearly showed that soda lime waste, borosilicate waste glass and composite of soda lime waste glass/MWCNTs are effective for the removal of lead (II) ions from waste water.
Determination of micronutrients, heavy metals and Proximate analysis of selected indigenous vegetables in Kirinyaga East Sub-County, Kirinyaga County
(Chuka University, 2025) Maina Janet Njeri
Socio-economic changes that have taken place in Africa have influenced peoples eating habits in both rural and urban set-ups. Indigenous vegetables are important for food insecurity, malnutrition reduction and therapeutics in sub-Saharan African countries. In Kirinyaga County, indigenous vegetables are underutilized or neglected due to some nutritional content are known and others unknown. Limited data exist on the precise levels of key micronutrients (e.g. iron, zinc, magnesium) in specific indigenous vegetables cultivated in Kirinyaga East Sub-County. Variability in nutrient content due to farming practices, soil types, and climatic conditions in the region has not been comprehensively studied. Leaves are the most preferred parts of indigenous vegetables for consumption. Although they have nutritional benefits, there is a need to determine the safety levels due to toxic metals in vegetables contaminated with pesticides, heavy metals and toxins leading to failure of certain organs of the human body. The study aimed to determine levels of micronutrients (calcium, iron, magnesium and zinc), heavy metals (lead and cadmium) using AAS (atomic absorption spectrometer) using and proximate analysis (ash content, moisture content, crude fat, protein content, carbohydrates and crude fiber) of selected African indigenous vegetables (African nightshade, Spider plant, Vine spinach and Pumpkin leaves) in Kirinyaga east sub county. The results revealed that African indigenous vegetables are rich in essential micronutrients. Pumpkin leaves had the highest calcium levels (14,070.81 mg/kg), spider plant showed the highest iron concentration (233.53 mg/kg), African nightshade recorded the highest zinc content (483.33 mg/kg), while vine spinach was richest in magnesium (570.87 mg/kg). Proximate analysis indicated that spider plant and African nightshade contained higher protein levels (up to 17.9%), while moisture content ranged from 82–91%. Crude fat levels were consistently low, confirming African indigenous vegetables as nutrient-dense but energy-light vegetables Cadmium and lead concentrations varied across species and locations, with African nightshade and vine spinach showing relatively higher levels. However, all values remained within WHO/FAO permissible limits, indicating that the vegetables are safe for human consumption in terms of heavy metal contamination. African indigenous vegetables have high nutritional potential and can be promoted as affordable dietary sources of essential minerals and protein to combat malnutrition.
Determination of Physico-Chemical Parameters and Removal of Bis(2-Ethylhexyl) Phthalate from Wastewater Using Prosopis juliflora Biochar/Carbon Nanotubes Composite Adsorbent
(Chuka University, 2024) Mutua John Mutinda
Phthalates, such as BEHP, are endocrine-disrupting compounds commonly used as plasticizers. Their presence in wastewater, often from industrial and household effluents, poses health risks including congenital anomalies, cancer, and chronic toxicity. The high cancer prevalence in Meru County has been linked to toxicants in effluent released into Kathita River, used for domestic purposes and irrigation. This is attributed to the inefficiency of the lagoon wastewater treatment technology in removing chemicals like BEHP. This study evaluates the physico-chemical properties and BEHP levels in wastewater from Meru Sewage Treatment Plant and explores the adsorption of BEHP using a Prosopis juliflora biochar/carbon nanotubes composite adsorbent. The temperature, pH, conductivity, turbidity, TDS, TSS, BOD and COD of the wastewater were determined using the standard APHA methods for wastewater, the concentrations of heavy metals using AAS while those of BEHP using HPLC. The following mean values were reported after data analysis: BEHP 0.055mg/L, Cu ND, Pb 0.042mg/L, Cd 0.0019 mg/L, COD 65.99 mg/L, TSS 29.3mg/L, TDS 639.17mg/L, turbidity 117.9FTU, conductivity 1079.9µS, pH 7.3, temperature 26oC and BOD5 65.9 mg/L; for dry season. Only pH and COD exceeded WHO limits for wastewater discharge into environment. The wet season parameter mean values were: COD 359.7mg/L, TSS 198.3mg/L, TDS 2094.2mg/L, turbidity 105FTU, conductivity 1244µS, pH 8, temperature 26 , BEHP 0.0429mg/L, Cu 0.47mg/L, Pb 0.037mg/L, Cd 0.056mg/L and BOD5 71.2mg/L. The temperature, TDS, BOD, Cu and Pb met WHO limit while conductivity, turbidity, TSS, COD and Cd exceeded. The composite adsorbent was characterized using FTIR and powder XRD. The dominant functional groups of the composite were C=O, CO2, OH-, Si-OH, C=N, MgO, CaCO3, and SiO2. The composite adsorbent was very efficient in the adsorption of BEHP with up to 96% removal in the samples at determined optimum adsorption parameters of; pH 5, temperature 24 , 15 minutes contact time and an adsorbent dose of 200 mg. The isotherm studies showed that the adsorption process was in agreement with the Freundlich isotherms with R2 value of 0.90469 while the kinetic studies revealed that BEHP adsorption followed pseudo second order model with R2 of 0.994. It was concluded that the application of biochar/CNTs composite adsorbent for the removal of BEHP from the wastewater is relatively cheaper and eco-friendly and should be applied for treatment of wastewater for irrigation and domestic use to improve water quality and minimize health risks associated with BEHP.
Determination of physicochemical parameters and estriol levels in nyeri waste water treatment plant and the adsorption of estriol using sugarcane bagassemultiwalled carbon nanotube composite
(Chuka University, 2024) Njue Jediel Mwenda
Persistent organic pollutants and endocrine disrupting substances have been found to be highly resistant to degradation. Exposure to these chemicals interferes with normal functioning of endocrine system by causing adverse effects such as cancer and impaired neurodevelopment. Waste water treatment plant mostly concentrate on removal of microorganisms leaving behind micro contaminants such as Estriol. The objective of the study was determination of physicochemical parameters and Estriol levels in Nyeri sewage treatment plant and the removal of Estriol using carbon nanotubes-sugarcane bagasse composite adsorbate. The composite was synthesized by oxidizing Multiwalled carbon nanotube in a ratio 3:1 nitric (V) acid to sulphuric (VI) acid mixture. Multiwalled carbon nanotube were then washed with deionized water and then added to a suspension of sugarcane bagasse powder in a ratio of 100:1 ,100:0.5 and 100:0.1. The characterization of the composites was done using Fourier transform infrared spectroscopy and X-ray diffraction techniques. The results showed that turbidity was higher in the dry season at 103.954 NTU compared to 46.5647 NTU in the wet season. Cd was recorded at 0.0168 mg/L during the wet season, which is significantly higher than the 0.0015 mg/L observed in dry season. Copper exhibited a concentration of 0.0833 mg/L in the wet season, which increased substantially to 1.0172 mg/L in the dry season. For Lead, the wet season value was 0.0159 mg/L, while the dry season value of 0.0050 mg/L. Dissolved Oxygen decreased markedly from 7.0617 mg/L in the wet season to 2.2396 mg/L in dry season. These differences in Pb, Cu, and DO levels were confirmed to be statistically significant p < 0.05 based on the Fisher’s LSD test with Bonferroni correction for p-value adjustment. Electrical Conductivity increased from 695.782 µS/cm in the wet season to 1056.104 µS/cm in the dry season. A t-test was conducted to compare Estriol concentrations between the wet and dry seasons at the Nyeri Water Treatment Plant. The mean Estriol concentration during the dry season was 1.444 ±0.671 Mg/L while in the wet season it was 0.982 ±0.870 Mg/L. The mean difference between the two seasons was 0.463, with a weighted standard deviation of 0.777. The optimum conditions obtained from the adsorption of Estriol were; equilibrium time 10 minutes, pH 4, maximum adsorbate concentration adsorbed at 0.1 g of composite 2 was 8 ppm. The composite that gave the best results was composite 2 (100:0.5) with 76.7% adsorption efficiency. The kinetic analysis of estriol best fitted the pseudo first order model with the rate constant for reaction as k1=−0.1−1, with an R2 value of 1.0. The analysis of adsorption isotherms for Estriol indicated that the Langmuir model provided a good model for fitting adsorption data than the Freundlich model, with a higher R² value of 0.5280 compared to 0.439 in Freundlich model. The enthalpy change (ΔH) for the process was calculated to be 26,165.44 J/mol, indicating that the adsorption is endothermic, meaning it absorbs heat from its surroundings. The entropy change (ΔS) was positive 68.85 J/mol-K suggesting an increase in disorder at the solid-liquid interface during adsorption. The Gibbs free energy change (ΔG) was negative at all the tested temperatures showing that the process naturally progresses without the need for external energy input.The findings demonstrated that sugarcane bagasse-carbon nanotube composite is a good low cost and environmentally friendly adsorbent for removal of Estriol from waste water.

Browse

Browsing Faculty of Science, Engineering and Technology (FSET) by Title