Geoffrey Mariga is a committed Christian who holds a Doctor of Philosophy in Information Technology degree Jomo Kenyatta University of Agriculture and Technology (JKUAT), M.Sc., Information Systems from the University of Nairobi and B.Sc. Mathematics
International Journal of Computer and Information Technology(2279-0764)
The paper presents feature extraction methods and classification algorithms used to classify maiz... more The paper presents feature extraction methods and classification algorithms used to classify maize leaf disease images. From maize disease images, features are extracted and passed to the machine learning classification algorithm to identify the possible disease based on the features detected using the feature extraction method. The maize disease images used include images of common rust, leaf spot, and northern leaf blight and healthy images. An evaluation was done for the feature extraction method to see which feature extraction method performs best with image classification algorithms. Based on the evaluation, the outcomes revealed Histogram of Oriented Gradients performed best with classifiers compared to KAZE and Oriented FAST and rotated BRIEF. The random forest classifier emerged the best in terms of image classification, based on four performance metrics which are accuracy, precision, recall, and F1-score. The experimental outcome indicated that the random forest had 0.74 ac...
International Journal of Computer and Information Technology(2279-0764), 2021
Recommender systems have taken over user’s choice to choose the items/services they want from onl... more Recommender systems have taken over user’s choice to choose the items/services they want from online markets, where lots of merchandise is traded. Collaborative filtering-based recommender systems uses user opinions and preferences. Determination of commonly used attributes that influence preferences used for prediction and subsequent recommendation of unknown or new items to users is a significant objective while developing recommender engines. In conventional systems, study of user behavior to know their dis/like over items would be carried-out. In this paper, presents feature selection methods to mine such preferences through selection of high influencing attributes of the items. In machine learning, feature selection is used as a data pre-processing method but extended its use on this work to achieve two objectives; removal of redundant, uninformative features and for selecting formative, relevant features based on the response variable. The latter objective, was suggested to i...
International Journal of Computer Applications Technology and Research
Computer vision has gained momentum in medical imaging tasks. Deep learning and Transfer learning... more Computer vision has gained momentum in medical imaging tasks. Deep learning and Transfer learning are some of the approaches used in computer vision. The aim of this research was to do a comparative study of deep learning and transfer learning in the detection of diabetic retinopathy. To achieve this objective, experiments were conducted that involved training four state-of-the-art neural network architectures namely; EfficientNetB0, DenseNet169, VGG16, and ResNet50. Deep learning involved training the architectures from scratch. Transfer learning involved using the architectures which are pre-trained using the ImageNet dataset and then fine-tuning them to solve the task at hand. The results show that transfer learning outperforms learning from scratch in all three models. VGG16 achieved the highest accuracy of 84.12% in transfer learning. Another notable finding is that transfer learning is able to not only achieve high accuracy with very few epochs but also starts higher than deep...
Kenya's Vision 2030 recognizes the enabling role of Information and Communication Technology ... more Kenya's Vision 2030 recognizes the enabling role of Information and Communication Technology (ICT) and anchors some of its key aspirations upon the availability and adoption of computers for schools. The overall objective of this strategy paper is to provide direction for acquisition of equitable and efficient laptop computers to all pupils in public primary schools. The existing education policy on ICT is imbedded in three documents namely; e-Government Strategy, National ICT Policy and Sessional Paper No. 1 of 2005 which is a Policy Framework for Education, Training and Research. There is a need therefore to consolidate these documents into one. The overall objective of the consolidation is to merge and integrate education policy on ICT including the scope, usage, administration and ways to address innovations and attendant Intellectual Property Rights. In the process of strategic planning for utilizing ICT in education, key stakeholders require to be consulted. In addition to...
Fixing failed computer programs involves completing two fundamental debugging tasks: first, the p... more Fixing failed computer programs involves completing two fundamental debugging tasks: first, the programmer has to reproduce the failure; second, s/he has to find the failure cause. Software debugging is the process of locating and correcting erroneous statements in a faulty program as a result of testing. It is extremely time consuming and very expensive. The term debugging collectively refers to fault localization, understanding and correction. Automated tools to locate and correct the erroneous statements in a program can significantly reduce the cost of software development and improve the overall quality of the software. This paper discusses fault localization, program slicing and delta debugging techniques. It identifies statistical fault localization tools such as Tarantula, GZoltar and others such as dbx and Microsoft Visual C++ debugger that provides a snapshot of the program state at various break points along an execution path. In conclusion we note that most software development companies spend a huge amount of resources in testing and debugging. A lot more research need to be conducted to fully automate the debugging process thereby reducing software production cost, time and improve quality.
Wambugu, Geoffrey Mariga http://erepository.uonbi.ac.ke:8080/xmlui/handle/123456789/10257 Date: 2... more Wambugu, Geoffrey Mariga http://erepository.uonbi.ac.ke:8080/xmlui/handle/123456789/10257 Date: 2012 Organizational Implementation of Information Systems Innovations (OIISI) Framework was developed in the context of Universities in Kenya and can be used to understand the implementation of Information Systems (IS) Innovations in Higher Education Institutions (HEIs). The aim of this study was to determine the degree of associations and relationships in the OIISI framework in HEIs and, in so doing, provided researchers and practitioners with a valid and reliable instrument that covered all the key constructs identified by the framework. In this study, the framework was tested in the context of HEIs in Kenya. To do so, data was collected from identified respondents in some selected HEIs that have implemented IS or were in the implementation process, analyzed and the outcomes presented, thereby validating the relationships. Judgmental and convenience sampling design was used to select HE...
One challenging issue in application of Latent Dirichlet Allocation (LDA) is to select the optima... more One challenging issue in application of Latent Dirichlet Allocation (LDA) is to select the optimal number of topics which must depend on both the corpus itself and user modeling goals. This paper presents a topic selection method which models the minimum perplexity against number of topics for any given dataset. The research set up scikit-learn and graphlab on jupyter notebook in the google cloud compute engine’s custom machine and then developed python code to manipulate selected existing datasets. Results indicate that the graph of perplexity against number of topics (K) has a strong quadratic behaviour around a turning point and opens upwards. This means that the graph has a minimum perplexity point that optimizes K. The paper presents a model of the optimum K in an identified interval and guides the calculation of this value of K within three iterations using quadratic approach and differential calculus. This improves inferential speed of number of topics and hyper parameter alp...
International Journal of Computer and Information Technology(2279-0764)
Deep learning has proven to be a landmark computing approach to the computer vision domain. Hence... more Deep learning has proven to be a landmark computing approach to the computer vision domain. Hence, it has been widely applied to solve complex cognitive tasks like the detection of anomalies in surveillance videos. Anomaly detection in this case is the identification of abnormal events in the surveillance videos which can be deemed as security incidents or threats. Deep learning solutions for anomaly detection has outperformed other traditional machine learning solutions. This review attempts to provide holistic benchmarking of the published deep learning solutions for videos anomaly detection since 2016. The paper identifies, the learning technique, datasets used and the overall model accuracy. Reviewed papers were organised into five deep learning methods namely; autoencoders, continual learning, transfer learning, reinforcement learning and ensemble learning. Current and emerging trends are discussed as well.
International Journal of Engineering and Advanced Technology
This paper provides an Extended Client Based Technique (ECBT) that performs classification on ema... more This paper provides an Extended Client Based Technique (ECBT) that performs classification on emails using the Bayessian classifier that attain in-depth defense by performing textual analysis on email messages and attachment extensions to detect and flag snooping emails. The technique was implemented using python 3.6 in a jupyter notebook. An experimental research method on a personal computer was used to validate the developed technique using different metrics. The validation results produced a high acceptable percentage rate based on the four calculated validation metrics indicating that the technique was valid. The cosine of similarity showed a high percentage rate of similarity between the validation labels indicating that there is a high rate of similarity between the known and output message labels. The direction for further study on this paper is to conduct a replica experiments, which enhances the classification and flagging of the snooped emails using an advanced classifica...
International Journal of Science and Engineering Applications, 2021
The rising number of malicious threats on computer networks and Internet services owing to a larg... more The rising number of malicious threats on computer networks and Internet services owing to a large number of attacks makes the network security be at incessant risk. One of the predominant network attacks that poses distressing threats to networks security are the brute force attacks. A brute force attack uses a trial and error algorithm to decode encrypted data such as passwords or Data Encryption Standard keys, through exhaustive effort (using brute force) rather than using intellectual strategies. Brute force attacks resemble legitimate network traffic, making it difficult to defend an organization that rely mainly on perimeter-based security solutions a major challenge. For stopping the occurrence of such attacks, several curable steps must be taken. This paper proposes an efficient mechanism for SSH-Brute force network attacks detection based on a supervised deep learning algorithm, Convolutional Neural Network. The model performance was compared with experimental results from 5 classical machine learning algorithms including Naive Bayes, Logistic Regression, Decision Tree, k-Nearest Neighbour, and Support Vector Machine. Four standard metrics namely, Accuracy, Precision, Recall, and the F-measure were used. Results show that the CNN-based model is superior to the traditional machine learning methods with 94.3% accuracy, a precision rate of 92.5%, recall rate of 97.8% and F1-score of 91.8% in terms of the ability to detect SSH-Brute force attacks.
International Journal of Computer and Information Technology(2279-0764)
The paper presents feature extraction methods and classification algorithms used to classify maiz... more The paper presents feature extraction methods and classification algorithms used to classify maize leaf disease images. From maize disease images, features are extracted and passed to the machine learning classification algorithm to identify the possible disease based on the features detected using the feature extraction method. The maize disease images used include images of common rust, leaf spot, and northern leaf blight and healthy images. An evaluation was done for the feature extraction method to see which feature extraction method performs best with image classification algorithms. Based on the evaluation, the outcomes revealed Histogram of Oriented Gradients performed best with classifiers compared to KAZE and Oriented FAST and rotated BRIEF. The random forest classifier emerged the best in terms of image classification, based on four performance metrics which are accuracy, precision, recall, and F1-score. The experimental outcome indicated that the random forest had 0.74 ac...
International Journal of Computer and Information Technology(2279-0764), 2021
Recommender systems have taken over user’s choice to choose the items/services they want from onl... more Recommender systems have taken over user’s choice to choose the items/services they want from online markets, where lots of merchandise is traded. Collaborative filtering-based recommender systems uses user opinions and preferences. Determination of commonly used attributes that influence preferences used for prediction and subsequent recommendation of unknown or new items to users is a significant objective while developing recommender engines. In conventional systems, study of user behavior to know their dis/like over items would be carried-out. In this paper, presents feature selection methods to mine such preferences through selection of high influencing attributes of the items. In machine learning, feature selection is used as a data pre-processing method but extended its use on this work to achieve two objectives; removal of redundant, uninformative features and for selecting formative, relevant features based on the response variable. The latter objective, was suggested to i...
International Journal of Computer Applications Technology and Research
Computer vision has gained momentum in medical imaging tasks. Deep learning and Transfer learning... more Computer vision has gained momentum in medical imaging tasks. Deep learning and Transfer learning are some of the approaches used in computer vision. The aim of this research was to do a comparative study of deep learning and transfer learning in the detection of diabetic retinopathy. To achieve this objective, experiments were conducted that involved training four state-of-the-art neural network architectures namely; EfficientNetB0, DenseNet169, VGG16, and ResNet50. Deep learning involved training the architectures from scratch. Transfer learning involved using the architectures which are pre-trained using the ImageNet dataset and then fine-tuning them to solve the task at hand. The results show that transfer learning outperforms learning from scratch in all three models. VGG16 achieved the highest accuracy of 84.12% in transfer learning. Another notable finding is that transfer learning is able to not only achieve high accuracy with very few epochs but also starts higher than deep...
Kenya's Vision 2030 recognizes the enabling role of Information and Communication Technology ... more Kenya's Vision 2030 recognizes the enabling role of Information and Communication Technology (ICT) and anchors some of its key aspirations upon the availability and adoption of computers for schools. The overall objective of this strategy paper is to provide direction for acquisition of equitable and efficient laptop computers to all pupils in public primary schools. The existing education policy on ICT is imbedded in three documents namely; e-Government Strategy, National ICT Policy and Sessional Paper No. 1 of 2005 which is a Policy Framework for Education, Training and Research. There is a need therefore to consolidate these documents into one. The overall objective of the consolidation is to merge and integrate education policy on ICT including the scope, usage, administration and ways to address innovations and attendant Intellectual Property Rights. In the process of strategic planning for utilizing ICT in education, key stakeholders require to be consulted. In addition to...
Fixing failed computer programs involves completing two fundamental debugging tasks: first, the p... more Fixing failed computer programs involves completing two fundamental debugging tasks: first, the programmer has to reproduce the failure; second, s/he has to find the failure cause. Software debugging is the process of locating and correcting erroneous statements in a faulty program as a result of testing. It is extremely time consuming and very expensive. The term debugging collectively refers to fault localization, understanding and correction. Automated tools to locate and correct the erroneous statements in a program can significantly reduce the cost of software development and improve the overall quality of the software. This paper discusses fault localization, program slicing and delta debugging techniques. It identifies statistical fault localization tools such as Tarantula, GZoltar and others such as dbx and Microsoft Visual C++ debugger that provides a snapshot of the program state at various break points along an execution path. In conclusion we note that most software development companies spend a huge amount of resources in testing and debugging. A lot more research need to be conducted to fully automate the debugging process thereby reducing software production cost, time and improve quality.
Wambugu, Geoffrey Mariga http://erepository.uonbi.ac.ke:8080/xmlui/handle/123456789/10257 Date: 2... more Wambugu, Geoffrey Mariga http://erepository.uonbi.ac.ke:8080/xmlui/handle/123456789/10257 Date: 2012 Organizational Implementation of Information Systems Innovations (OIISI) Framework was developed in the context of Universities in Kenya and can be used to understand the implementation of Information Systems (IS) Innovations in Higher Education Institutions (HEIs). The aim of this study was to determine the degree of associations and relationships in the OIISI framework in HEIs and, in so doing, provided researchers and practitioners with a valid and reliable instrument that covered all the key constructs identified by the framework. In this study, the framework was tested in the context of HEIs in Kenya. To do so, data was collected from identified respondents in some selected HEIs that have implemented IS or were in the implementation process, analyzed and the outcomes presented, thereby validating the relationships. Judgmental and convenience sampling design was used to select HE...
One challenging issue in application of Latent Dirichlet Allocation (LDA) is to select the optima... more One challenging issue in application of Latent Dirichlet Allocation (LDA) is to select the optimal number of topics which must depend on both the corpus itself and user modeling goals. This paper presents a topic selection method which models the minimum perplexity against number of topics for any given dataset. The research set up scikit-learn and graphlab on jupyter notebook in the google cloud compute engine’s custom machine and then developed python code to manipulate selected existing datasets. Results indicate that the graph of perplexity against number of topics (K) has a strong quadratic behaviour around a turning point and opens upwards. This means that the graph has a minimum perplexity point that optimizes K. The paper presents a model of the optimum K in an identified interval and guides the calculation of this value of K within three iterations using quadratic approach and differential calculus. This improves inferential speed of number of topics and hyper parameter alp...
International Journal of Computer and Information Technology(2279-0764)
Deep learning has proven to be a landmark computing approach to the computer vision domain. Hence... more Deep learning has proven to be a landmark computing approach to the computer vision domain. Hence, it has been widely applied to solve complex cognitive tasks like the detection of anomalies in surveillance videos. Anomaly detection in this case is the identification of abnormal events in the surveillance videos which can be deemed as security incidents or threats. Deep learning solutions for anomaly detection has outperformed other traditional machine learning solutions. This review attempts to provide holistic benchmarking of the published deep learning solutions for videos anomaly detection since 2016. The paper identifies, the learning technique, datasets used and the overall model accuracy. Reviewed papers were organised into five deep learning methods namely; autoencoders, continual learning, transfer learning, reinforcement learning and ensemble learning. Current and emerging trends are discussed as well.
International Journal of Engineering and Advanced Technology
This paper provides an Extended Client Based Technique (ECBT) that performs classification on ema... more This paper provides an Extended Client Based Technique (ECBT) that performs classification on emails using the Bayessian classifier that attain in-depth defense by performing textual analysis on email messages and attachment extensions to detect and flag snooping emails. The technique was implemented using python 3.6 in a jupyter notebook. An experimental research method on a personal computer was used to validate the developed technique using different metrics. The validation results produced a high acceptable percentage rate based on the four calculated validation metrics indicating that the technique was valid. The cosine of similarity showed a high percentage rate of similarity between the validation labels indicating that there is a high rate of similarity between the known and output message labels. The direction for further study on this paper is to conduct a replica experiments, which enhances the classification and flagging of the snooped emails using an advanced classifica...
International Journal of Science and Engineering Applications, 2021
The rising number of malicious threats on computer networks and Internet services owing to a larg... more The rising number of malicious threats on computer networks and Internet services owing to a large number of attacks makes the network security be at incessant risk. One of the predominant network attacks that poses distressing threats to networks security are the brute force attacks. A brute force attack uses a trial and error algorithm to decode encrypted data such as passwords or Data Encryption Standard keys, through exhaustive effort (using brute force) rather than using intellectual strategies. Brute force attacks resemble legitimate network traffic, making it difficult to defend an organization that rely mainly on perimeter-based security solutions a major challenge. For stopping the occurrence of such attacks, several curable steps must be taken. This paper proposes an efficient mechanism for SSH-Brute force network attacks detection based on a supervised deep learning algorithm, Convolutional Neural Network. The model performance was compared with experimental results from 5 classical machine learning algorithms including Naive Bayes, Logistic Regression, Decision Tree, k-Nearest Neighbour, and Support Vector Machine. Four standard metrics namely, Accuracy, Precision, Recall, and the F-measure were used. Results show that the CNN-based model is superior to the traditional machine learning methods with 94.3% accuracy, a precision rate of 92.5%, recall rate of 97.8% and F1-score of 91.8% in terms of the ability to detect SSH-Brute force attacks.
Uploads
Papers by Geoffrey Wambugu