Master of Science in Computer Information Systems

Deep Learning Approach for Detection and Prediction of Pest Infections On Plants in Greenhouses

2025-09-01T00:00:00Z

Deep Learning Approach for Detection and Prediction of Pest Infections On Plants in Greenhouses Sambu, Bridgite Mueni Pest infestations remain a serious threat to greenhouse agriculture ultimately resulting in reduced yields, increased cost of production, and food safety concerns. Traditional pest monitoring strategies are predominantly manual in nature, which can be laborious, time-consuming, and exhibit human error. These downfalls can affect timely action on issues and a positive relationship management with pest monitoring can mean excessive pesticide use with financial and/or environmental impacts. This paper proposed an AI-enabled hybrid deep learning model to automate pest detection and estimate the probability of a pest outbreak. The hybrid model combined Convolutional Neural Networks (CNNs) for spatial (image-based) pest identification with Long Short-Term Memory (LSTM) networks to estimate probabilities of outbreaks based on sequences of environmental ((local) environmental data temperature, humidity, as well as, recorded counts of pests) and pest counts to help improve timing of interventions and proactive pest management solutions. The study made use of both primary and secondary datasets. The primary dataset was collected from three greenhouses in Limuru, Naivasha, and Thika, Kenya as well as high-resolution crop images (48 mega-pixels) representing various pest infections. Local environmental data for temperature, humidity, and pest counts were collected from the greenhouses to use as sequential variables in the LSTM component of the workflow and assist with estimating the accuracy of predicted outbreaks. Secondary datasets of pre-annotated pest images and historical climate records, including PlantVillage and IP102, provided bulked training data that improved models’ robustness and generalizability to unseen datasets. The researcher ensured stratified sampling to capture representation of all greenhouse types, farm sizes, and agro-climatic conditions. As part of preprocessing, all datasets underwent the following steps: image augmentation, noise removal and feature normalization. All model training, including hyperparameter tuning, occurred in a GPU-enabled Google Colab environment, and early stopping was used to avoid overfitting. The hybrid CNN-LSTM model produced a classification accuracy of 94.7%, precision of 93.9%, recall of 93.8%, and F1-score of 93.2%. The time-series predictive component of the LSTM model produced strong predictive performance with a mean absolute error (MAE) of 0.14, and R² value of 0.89. This showed that the environmental and measured sequence data of pests improve outbreak prediction. It was shown that the hybrid model accurately identifies pest infections, as well as predicting their outbreaks, which will contribute to providing early and timely interventions in the pest control system. Both high-resolution image data and measurements of local environments support a scalable and resilient choice for greenhouse pest management - the use of less pesticides, leading to sustainability in agriculture. In conclusion, this study has demonstrated the effectiveness of a hybrid deep learning framework for integrated pest management with operational and financial implications.

Development of A Machine Learning-Based Model Using a Decision Tree for Detecting Fake News: Analyzing Techniques for Accurate Content Verification

2025-09-01T00:00:00Z

Development of A Machine Learning-Based Model Using a Decision Tree for Detecting Fake News: Analyzing Techniques for Accurate Content Verification Tomba, Kinkosi Esther With the growing spread of information through social media and online news, identifying fake news has become increasingly important. To explore this issue, the Pew Research Center conducted surveys in the U.S. to examine how adults access news on social platforms, focusing on the behaviors and demographics of users who rely on these channels. This research sought to address a major gap in traditional fake news detection approaches, which were largely manual and lacked the sophistication of advanced machine learning and AI methods. Such conventional techniques struggle to handle the complexity and contextual manipulation of information, where accurate facts can be framed misleadingly. To overcome these shortcomings, the study developed a machine learning–based model to detect fake news by analyzing article content and recognizing patterns of misinformation. It utilized advanced natural language processing (NLP) techniques and supervised learning algorithms. Decision Trees (achieving 99.67% accuracy), Logistic Regression (99.13%), and Random Forest (99.15%). Processes such as tokenization and TF-IDF were applied to train the model on the ISO Fake News dataset, which combined real news from Reuters.com with fake news from unreliable sources flagged by PolitiFact and Wikipedia. Model performance was evaluated using accuracy, precision, recall, and F1-score, all reaching 99.67%, demonstrating exceptional detection capability. This work contributes to the field of machine learning by enhancing NLP methods and improving the effectiveness of fake news detection models. Future research is encouraged to expand datasets, incorporate multiple languages, employ deep learning like RNNs, CNNs, and Transformers (e.g., BERT and RoBERTa) for richer contextual understanding, and establish benchmarks based on real-world case studies.

Integrating Ai Tools for Real-Time Anomaly Detection in Cloud Vpns: A Case of Owncloud.

2025-09-01T00:00:00Z

Integrating Ai Tools for Real-Time Anomaly Detection in Cloud Vpns: A Case of Owncloud. Boyani, Momanyi Zipporah The increasing reliance on cloud-based Virtual Private Networks (VPNs) has significantly improved the security and scalability of digital infrastructures. However, the increasing complexity of these systems introduced new challenges in ensuring their security, particularly in detecting anomalies such as unauthorized access, abnormal traffic, and data breaches in real time. This research addressed the problem of inadequate anomaly detection in the dynamic cloud VPN environments by investigating the integration of Artificial Intelligence (AI) tools to enhance real-time threat detection, using OwnCloud as a case study. The research aimed to identify effective AI models for anomaly detection, develop a real-time AI-based prototype, and evaluate its performance in detecting anomalies within cloud VPN traffic. Both supervised and unsupervised machine learning techniques were explored, including Isolation Forest and Long Short-Term Memory (LSTM) models. Simulated VPN traffic data was generated using Mininet, and Apache Kafka was employed to stream this data in real time to a Spark-based AI detection engine. Anomaly detection outputs were logged and visualized using the Kibana dashboard, while alerts were configured to trigger based on spikes and deviations from normal traffic patterns. The prototype demonstrated the feasibility of AI-based tools in identifying unknown and evolving threats more effectively than traditional signature-based systems. Unlike conventional methods that rely on historical data and static thresholds, the AI-driven system adapted to emerging threat patterns and significantly reduced false positives. The comparative analysis of AI models confirmed that the hybrid (LSTM + Isolation Forest) model was the most effective AI-based approach for anomaly detection in simulated cloud VPN traffic. It not only delivered superior performance metrics but also demonstrated adaptability in real-world scenarios where labeled anomalies are scarce, and encrypted traffic restricts payload inspection. The model recorded the highest Precision of 0.94, Recall of 0.91, F1-Score of 0.92, and Accuracy of 0.93. The developed AI-based prototype system effectively achieved real-time anomaly detection in OwnCloud VPN traffic. Its hybrid architecture, based on LSTM and IF, delivered accurate, timely, and interpretable results, hence validating its potential for integration into real-world cloud security systems. The ROC curve for the real-time anomaly detection prototype revealed exceptional performance, with an AUC score of 0.98 confirming its effectiveness in distinguishing between normal and anomalous traffic. The findings highlighted the potential of AI to improve the responsiveness and accuracy of intrusion detection mechanisms in cloud-based environments. In conclusion, the research successfully demonstrated that AI tools can enhance real-time anomaly detection in the cloud VPNs, offering improved threat response and reduced false alarms. This research will add to the existing knowledge of AI integration to improve the security of cloud VPNs by exploring a case study in real life. It is recommended that future implementations expand on this approach by integrating more advanced deep learning models, refining real-time alert systems, and applying the solution to diverse cloud platforms to further validate scalability and robustness.

Enhancing Diabetes Classification Using A Weighted Ensemble Of Tabnet, Xgboost And Random Forest

2025-09-01T00:00:00Z

Enhancing Diabetes Classification Using A Weighted Ensemble Of Tabnet, Xgboost And Random Forest Obunge, Duncan Ogindo The increasing prevalence of Diabetes mellitus (DM), a leading global health challenge, is significantly impacting the healthcare systems. Accurate and interpretable classification models are crucial in advancing early diagnosis and effective intervention. While traditional machine learning techniques like Extreme Gradient boosting (XGBoost) and RandomForest based models have demonstrated robust classification performance on tabular medical datasets, however, they have continued to face challenge of model interpretability. Deep learning models, like TabNet, cater the two-pronged benefits of feature selection learning and interpretability via attention mechanisms. This study developed a weighted ensemble model that combines TabNet, XGBoost and RandomForest based models to address the trade-off between interpretability and strong performance. The study utilized the Pima Indian Diabetes dataset as secondary data and expert clinical validation. The dataset, contained 768 tuples with 8 features, related to diabetes risk factors. The ensemble assigns optimized weights to the classifications of the three models, drawing on their complementary strengths. The results indicated that the weighted ensemble model outperformed the individual models; while preserving interpretability. The implementation achieved a balanced accuracy of 0.8630 ± 0.0146 (median 0.8350), precision of 0.8163 ± 0.0442 (median 0.8018), recall of 0.9376 ± 0.0341 (median 0.8900), F1 score of 0.8401 ± 0.0110 (median 0.8436), and ROC-AUC score of 0.9026 ± 0.0172 (median 0.9044), while the traditional machine learning models based on XGBoost attained (0.8103 ± 0.0270 (0.8150) balanced accuracy, 0.7888 ± 0.0287 (0.7890) precision) and RandomForest achieved (0.8060 ± 0.0250 (0.8100) balanced accuracy, 0.7451 ± 0.0312 (0.7768) precision) algorithms. Feature importance analysis revealed the top most significant predictors of diabetes based on normalized scores as; glucose level (≈1), followed by age (≈0.458), insulin (≈0.434) and body mass index(BMI) (≈0.13) hence providing valuable clinical insights. This research contributes a novel computational framework that leverages a weighted ensemble learning techniques while preserving model explainability; a critical advancement for healthcare-aligned machine learning systems. This methodological contribution extends beyond diabetes classification to potentially benefit various clinical decision support systems operating on limited-feature medical datasets.

M-Shopping Application’S Construct: Modelling Proximity and Route Map With Regard to Consumers’ M-Shopping Behaviour in Kenyas Nairobi Metropolitan Region

2024-09-01T00:00:00Z

M-Shopping Application’S Construct: Modelling Proximity and Route Map With Regard to Consumers’ M-Shopping Behaviour in Kenyas Nairobi Metropolitan Region NDOLO, AGBESTERS MWOVE Mobile shopping only relates to specific elements of the purchase process, mostly in business-to-consumer and consumer-to-consumer scenarios. M-shopping is one of the most popular online pastimes in the world, with e-commerce sales reaching 4.28 trillion US dollars globally in 2020 and revenues predicted to reach 5.4 trillion US dollars in 2022. Mobile commerce (M-commerce) diverges from traditional electronic commerce (E-commerce) due to disparities in its user interface and the correlated factors of risk, interactivity, ubiquity, localization services, and patterns of usage. Social media is influencing consumer purchase decisions more and more, overshadowing conventional opinions on goods and services. This study sought to assess proximity and route map with regard to consumers’ M-shopping behavior in Kenyas Nairobi Metropolitan Region. The objectives of the study were; to investigate place convenient modelled in the current m-shopping applications, to examine the influence of proximity and route map in the current m-shopping applications and to model proximity and route map in an m-shopping application. A cross-sectional research design was adopted and a sample size of 106 respondents determined using a simple random sampling technique. The primary data was collected through structured survey questionnaires, after which, STATA software was used for its analysis. The study established a statistically significant relationship between the convenient in m-shopping applications and consumer’s m-shopping behaviour (ꭕ2=6.370a, p=.041<0.05). A statistically significant relationship between the proximity in the current m-shopping applications and consumer’s m-shopping behaviour was also noted (ꭕ2=13.234a, p=.001<0.05). Further, it established a positive statistically significant relationship between the route map in the current m-shopping applications and consumer’s m-shopping behaviour (ꭕ2=72.192a, p=.000<0.05). Analyzing the influence of proximity distance on consumer’s behavior revealed that, consumers tend to favor businesses located closer to their current location when making buying decisions through m-shopping applications. Further, provision of vendor-consumer real-time navigation assistance would enhance consumer’s behavior. Despite the high deployment of mobile shopping applications, less attention is given to the proximity distance and route map aspects in the mobile shopping landscape. Consequently, this negatively impacts on the consumer’s m-shopping behavior. Besides, this study serves as an intervention target for improving the consumer’s behavior. It’s therefore recommendable that the developers should adopt a holistic and consumer-vendor centered approach to mobile applications. In addition, implementation studies should be employed to test this model and identify other factors that may be useful to effectively improve on the m-shopper’s behavior. Future research opportunities lie in exploring advanced technologies such as augmented reality (AR) integration with route maps in M-Shopping applications or analyzing data-driven approaches for personalized recommendations based on both proximity distance and historical buying patterns. By continuously evolving strategies informed by consumer behavior analysis within M-Shopping applications, businesses can stay ahead in meeting the dynamic needs of tech-savvy consumers in markets like Nairobi Metropolitan where digital innovation continues to drive shopping trends forward.

Assessment of the Effectiveness of Non-Technical Approach to Cyber Security Management for Kenya Ministry of Lands and Physical Planning Nlims System.

2024-10-01T00:00:00Z

Assessment of the Effectiveness of Non-Technical Approach to Cyber Security Management for Kenya Ministry of Lands and Physical Planning Nlims System. NDUNG’U, GABRIEL Cyber security threats and vulnerabilities are influenced by human behavior, making them complex systems. The conventional reductionist approach to managing technical aspects of cyber security is limited, as it cannot predict the non-technical aspects of cyber security. The complexity approach, however, offers a robust and effective way to predict social aspects of cyber security. This study aims to assess the effectiveness of the Ministry of Land's approach to non-technical cyber threats and vulnerabilities. Current studies focus predominantly on the technical aspect cyber security and less on social aspect of cyber security at the detriment of the social side so does the setup and implementation of information security system. In aggregate system that focus on the technical side are weak on the social aspect and predisposed to social-engineering attack, Data will be collected through face-to-face interviews with Ministry of land staff linked to the NLIMS system in Nairobi. The findings indicate that at least 70% of staff lack knowledge of social engineering attacks, their conduct, and skills to prevent or stop them. Lower-rank staff access information they are not authorized to access through the workstation resource sharing policy. The study reveals that KMLPP's non-technical approach to cyber security is ineffective in protecting sensitive information and preventing staff from accessing sensitive data through dumb star diving. It also highlights the vulnerability of lower-rank staff to shoulder surfing and workstation privacy exploits. The study proposes a socio-technical cybernetic enterprise model, which focuses on staff relationships and restricts information access without appropriate privileges, ensuring a more secure environment for NLIMS.

Deep Learning Network Intrusion Detection With the Conv1d-Lstm Model: Integrating Cnn and Lstm for Superior Performance

2024-09-01T00:00:00Z

Deep Learning Network Intrusion Detection With the Conv1d-Lstm Model: Integrating Cnn and Lstm for Superior Performance Lukogo, Cikambasi Ciza The escalating frequency and complexity of cyber-attacks present significant threats to corporate networks, resulting in financial losses, reputational harm, and possible data breaches. Traditional Intrusion Detection Systems (IDS), which rely on predefined signatures and rules, have proven inadequate in addressing these advanced threats due to high rates of both false positives and negatives. This inadequacy necessitates the development of more advanced intrusion detection methods. This thesis introduces a novel AI-based intrusion detection model leveraging deep learning techniques to enhance corporate network security. The proposed model utilizes convolutional neural networks (CNN) and recurrent neural networks (RNN) to analyze network traffic data from the comprehensive CSE-CIC-IDS-2018 dataset, which encompasses a wide array of attack types and provides a realistic representation of modern network traffic. By capturing complex patterns and temporal dependencies in the data, these deep learning algorithms are particularly effective in detecting sophisticated intrusion attempts A key contribution of this research is the development of a hybrid detection approach that fuses convolutional neural network and recurrent neural network algorithm. This hybrid model enhances detection accuracy and reduces false alarms. The model's performance is rigorously evaluated using metrics such as precision, recall, and F1 score, demonstrating superior detection capabilities with a 99.97% in precision, 99.95% in recall, 99.97% in accuracy, and 99.96% in F1-score outperforming the other models. To address challenges related to data quality, the study incorporates extensive data preprocessing steps, including feature selection, encoding, and scaling. The high computational demands of training deep learning models are mitigated using cloud-based resources. Furthermore, visualization strategies are used to enhance the model's interpretability, offering a glimpse into its decision-making process. The findings of this research have significant implications for network security administrators, researchers, educators, and policymakers. Network security administrators can apply these insights to enhance their defensive strategies against cyber threats. Researchers and educators can leverage the advanced methodologies presented in this study, while policymakers can utilize the findings to inform the development of more effective network security policies and standards. This research advances the field of cybersecurity by proposing and evaluating a novel AI-based intrusion detection model. It underscores the critical importance of integrating advanced AI methodologies into IDS frameworks to protect corporate networks from evolving cyber threats. By improving the accuracy and reliability of intrusion detection systems, this study contributes to the overall security of digital operations in organizations worldwide, highlighting the transformative potential of AI in contemporary cybersecurity.

Information Management Systems Strategies to Support The Integration of Kenya’S E-Government Services A Case Study of The Kenya Revenue Authority

2024-09-01T00:00:00Z

Information Management Systems Strategies to Support The Integration of Kenya’S E-Government Services A Case Study of The Kenya Revenue Authority Tejwok, Simon The purpose of this research thesis was to examine how information management system methods can be applied to successfully integrate electronic government service modules for effective service delivery and revenue collection. Despite reporting higher collections, the Kenya Revenue Authority (KRA) failed to meet its income objectives for the 2022–2023 fiscal year by Kes107 billion. The domestic excise tax was one of the bands that fell short of the goal, performing below average at 91.4 percent to achieve Kes 68.124 billion. However, the taxman claimed that earnings from the betting excise tax came to Sh6.65 billion, exceeding the objective of Kes 5.72 billion. This result was ascribed by KRA to the betting companies' smooth integration into the KRA tax system. This integration has greatly increased revenue collection and expedited the sector's tax payment procedure. The agency connected its systems with the betting companies' systems around the end of 2022. This connectivity has made it possible to collect taxes in real-time and has increased visibility into the revenue these businesses generate. This descriptive-analytical study used a descriptive research design to investigate the information systems management strategies for the integration of e-government services with the Kenya Revenue Authority as a case study. This was done on a targeted population of 256 management staff extracting a sample size of 72 respondents using structured questionnaires. Interviews were conducted on normal tax payers, corporate taxpayers from 2 select companies, 2 KRA commissioners related to the ICT integration strategy and 4 senior ICT staff attached to the ICT authority. The questionnaires received responses from 66 staff members in total, or almost 91% of the intended audience. IBM's SPSS version 26 was used to examine the data sets. The emergent independent variables (information systems, digital infrastructure and stakeholder involvement) and the dependent variable underwent qualitative data analysis. After that, inferential statistics were used to determine how information management system strategies would influence KRA system integration with the e-government platform. This included regression and correlation analysis. Information systems and stakeholder participation had a positive connection with the dependent variable. Digital infrastructure strategy did not have a significant correlation with integration of e-government services as most respondents did not agree with the compliance of standards and reliability of the current infrastructure at KRA (data center issues M=2.59, SD=1.268; Network infrastructure issues M=2.91, SD=0.944). In conclusion, the findings indicated that information systems strategy had a strong significance on the integration with e-government services despite the weak significance of the digital infrastructure variable. A proper information management system to check the flow of revenue data is necessary. This research focused on examining the utilization of information management systems integration strategies by the KRA.

AI Based Framework for Government Oversight of Personal Data Consent Compliance A Case Study Of Nairobi County

2024-09-01T00:00:00Z

AI Based Framework for Government Oversight of Personal Data Consent Compliance A Case Study Of Nairobi County VUNDI,MUSYOKA, GEOFFREY In the rapidly evolving digital landscape, protecting individual privacy and ensuring compliance with personal data regulations have become critical priorities. This study addresses the growing challenge of insufficient government oversight in monitoring real time compliance with personal data consent, with a focus on Nairobi County as a case study. It introduces an AI-based framework designed to automate the detection of privacy breaches, verify adherence to consent agreements, and strengthen regulatory enforcement processes. Grounded in regulatory compliance theory, the research aims to enhance the capacity of oversight bodies by utilizing AI technology to analyze vast datasets, improving the speed, accuracy, and efficiency of compliance monitoring. A mixed-methods research design was adopted, integrating both qualitative and quantitative approaches. Key stakeholders from prominent organizations such as Safaricom PLC, the Kenya Revenue Authority, Equity Bank Kenya, the Ministry of Information, Communications, and Technology (ICT), and the United Nations Office at Nairobi (UNON) were engaged. Data was collected through semi-structured questionnaires and in-depth interviews with government regulators and private sector representatives responsible for managing personal data. A purposive sampling method was employed, selecting 195 respondents to ensure a comprehensive and representative dataset. Data analysis involved thematic analysis for qualitative data and statistical techniques for quantitative data. Findings indicate that the AI-based framework significantly improves the detection and prevention of data privacy violations, optimizes compliance processes, and reduces reliance on manual oversight. Enhanced governance structures and heightened user awareness emerged as crucial factors in promoting better compliance. However, challenges such as regulatory adaptation and limited resources were identified. The study concludes that AI holds transformative potential for government oversight by increasing transparency, accountability, and operational efficiency. It recommends that regulatory bodies, particularly the Ministry of ICT, adopt AI-driven solutions and foster public-private partnerships to ensure effective, comprehensive data governance. This approach is vital for addressing emerging privacy challenges in a data-driven world

A Framework for Optimizing Pharmacy Inventory Management System Performance Using Cloud Computing and Machine Learning a Case Study of Nairobi County

2024-09-01T00:00:00Z

A Framework for Optimizing Pharmacy Inventory Management System Performance Using Cloud Computing and Machine Learning a Case Study of Nairobi County Chebet, Kelvin The purpose of this study was to address how pharmacy inventory management systems can be improved using cloud computing and machine learning. The main aim was to enhance efficacy, accuracy and efficiency in inventory management practices within the pharmaceutical sector. The problem identified was about the inefficiencies and challenges present in conventional stock control methods such as manual tracking mechanisms and outdated ones. Because of these inefficiencies, issues such as stock-outs, excesses, and lack of real-time information critical for decision-making processes arise. To overcome this challenge, a quantitative research design was used where data was collected through questionnaires and interviews from a diverse group of pharmacy personnel. The sample included public and private pharmacies in Nairobi County through stratified random sampling. The methodology involves the use of questionnaires for quantitative data collection on ongoing inventory management practices as well as technological readiness. This study expects that by utilizing cloud computing and machine learning algorithms there will be an inclusive framework created for optimizing pharmacy inventory management systems. The results indicated a need for the implementation of the proposed machine learning and cloud computing framework as the respondent indicated a high dissatisfaction it their current inventory management systems which were indicated to have major challenges that contributed to financial losses, customer dissatisfaction among other. Additionally, this research provides practical recommendations for implementing cloud computing platforms or machine learning solutions which could transform the traditional approach to inventory management thereby enhancing patient care outcomes.