Browse published research outputs

Aim: This study aimed to improve movie recommendation accuracy under sparse rating conditions by developing an ensemble system that uses dynamically weighted user profiles to generate personalised recommendations. Design: An experimental quantitative design was adopted. The ensemble recommendation system integrated Case Evaluation using K-means (CE-KM), Case Evaluation using Weighted Association Rule Mining (CE-WARM), and an Adaptive Naive Bayes Probabilistic technique (ADNB). Methodology: The framework comprised three stages: dataset preprocessing and user-profile construction, pattern recognition from users’ movie histories, and the generation and ranking of candidate recommendations. Nine weighted profile attributes represented user interactions and content preferences. The component methods were examined in five experiments using different K values. Accuracy and F-measure were used for evaluation, and the predictions from the three methods were combined to produce the top-ranked movies for an active user. Results: Within the reported experiments, CE-KM achieved its highest accuracy of 0.9323 at K = 70, followed by ADNB which achieved 0.8197 at K = 80 then, CE-WARM which achieved 0.7980 at K = 60. The methods generated different candidate lists, indicating that the profiling, association, and probabilistic components provided complementary outputs. The final ensemble ranking selected movies from the combined predictions. Conclusion: The framework demonstrates a structured approach to combining dynamic user profiling with multiple machine-learning techniques. However, its comparative performance under controlled sparsity conditions requires further evaluation.

Performance is an important attribute of a software system. The need to use suitable techniques to analyze the performance characteristics of any software system is paramount most especially at the architectural level. Performance as a requirement often originates from the organization’s business goals. In the case of the Made in Nigeria Primary Healthcare Information System (MINPHIS) as a product, the business goal is to enter new and emerging geographic markets where the system will be expected to perform effectively. In this paper we considered performance refinement issues of MINPHIS application from the architectural point of view. We present the entity-relationship diagram (ERD) as a role-based quality model of performance attributes showing the main stakeholder roles of MINPHIS software system. The performance goal refinement will be to support regulations that require life system like the MINPHIS application that keeps patient records and generates various reports for health management and research purposes. The paper also presents the module view of MINPHIS architecture, and show who the major stakeholders are and the performance refinement characterization.

Development of a Summarisation System for Web Pages is an MSc project that designed a summarisation system for search engine results by looking into the web pages and summarising it to a page. The purpose of this work is to provide information to the solution of multi web page summarisation problem by providing a system that will summarise two to four web pages at a time to easy researchers in getting right information from the internet at a minimal time. The work produces a temporary webpage that contains extract of salient sentences from pages got from query posed on Google search engine. Uniform Resource Locator (URL) data harvested from Google search engine was used to develop and test the summarisation system. The web page summarisation system was designed by first analysing and characterising web pages using Text difference function to check for relevancy of the web pages to be summarised. Pair wise bipartite graph theory was used to check for redundancy and score the sentences. The result from the system shows that time to glance through all web pages before getting concrete summary is reduced using this package.

Voice fraud has recently become a critical cybersecurity concern, where fraudsters impersonate trusted individuals over phone calls to deceive victims and obtain confidential information.Conventional fraud detection systems, which rely on predetermined rules or post-call detection, are increasingly becoming ineffective against evolving fraud patterns and do not ensure user safety in real-time.This study proposes the design and implementation of a realtime voice fraud detection system that analyses real-time conversations to identify potential fraudulent activity.The proposed system integrates OpenAI's Whisper model for speech-to-text transcription with a fine-tuned Phi-3-mini Large Language Model (LLM) for contextual fraud detection.The system architecture uses the Browser MediaStream API for audio capture and WebSocket communication to transmit streaming data to a FastAPI-based backend for real-time processing.The fine-tuned LLM was trained on a dataset comprising 447 labeled conversational samples and subsequently evaluated on 307 unseen test samples, achieving an overall accuracy of 95%.Experimental results demonstrate a precision of 97% and a recall of 98% for normal conversation, and a precision of 75% and a recall of 67% for fraudulent conversation.These findings underscore the effectiveness of leveraging context-aware language modelling and real-time audio transcription in detecting fraudulent speech patterns, thereby surpassing the limitations of traditional rule-based systems.The study contributes to the body of knowledge by demonstrating the feasibility of combining speech processing and large language models to enable proactive, real-time detection of voice fraud during phone calls.

Web spam detection is a critical issue in today’s rapidly growing usage of the Internet and the World Wide Web. The upsurge of web spam has significantly deteriorated the Quality of Services (QoS) of the World Wide Web. The degeneration of the quality of search engine results has given rise to researches on the detection of spam pages efficiently and accurately. Existing user-behaviour oriented web spam detection models employed the content-based, link-based and other features of webpages for classification of web spams. These user-behaviour techniques either implemented singly or combined has achieved good detection performance. However, the effectiveness of these features in identifying Web spams correctly needs to be determined. In this study, predictive web spam detection models that employed all related user-behaviour features of webpages were developed and evaluated. The content, link, and obvious-based features datasets were collected from an online repository. Relevant features were extracted using an improved Filter-based method. Six user-behaviour related features extracted from the datasets were used to combine the datasets to generate all possible subset of feature space required, such that 7 new datasets were generated for the study. Multi-Layer Perceptron (MLP) approach was adopted as a classifier for each of the identified features. Python Machine Learning Library was used to simulate the models using percentage splits of 60/40%, 70/30% and 80/20% ratio for training/testing dataset and the performances were evaluated using accuracy, True Positive (TP) rate, False Positive (FP) rate and precision as metrics. The result showed that for the majority of the datasets the formulated models have shown an increase in efficiency after feature selection. The MLP classifier was able to achieve the best result of 66.0% accuracy when the link-based dataset was used with feature selection. The study concluded that link-based features of a user is sufficient and effective for the detection of web spams. Keywords: Webspam, Content-based, Link-based, features, user-behaviour, evaluation DOI : 10.7176/NCS/10-07 Publication date :December 31 st 2019

This paper formulates, simulates and assess an improved data clustering algorithm for mining web documents with a view to preserving their conceptual similarities and eliminating the problem of speed while increasing accuracy. The improved data clustering algorithm was formulated using the concept of K-means algorithm. Real and artificial datasets were used to test the proposed and existing algorithm. The proposed algorithm was simulated using the fuzzy logic and statistical toolbox in Matlab 7.0. The simulated results were compared with the existing data clustering algorithm using accuracy, response time, adjusted rand index and entropy as performance parameters. The results show an improved data clustering algorithm with a new initialization method based on finding a set of medians extracted from a dimension with maximum variances. The results of the simulation showed that the accuracy is at its peak when the number of clusters is 3 and reduces as the number of clusters increases. When compared with existing algorithm, the proposed clustering algorithm showed an accuracy of 89.3% while the existing had an accuracy of 88.9%. The entropy was stable for both algorithms with a value of 0.2485 at k = 3. This also decreases as the number of clusters increase until when the number of clusters reached eight where it increased slightly. The adjusted rand index values varied from 0 to 1 for both clustering algorithms. The existing method achieved a value of 53% as compared with the proposed method which achieved an adjusted rand index value of 63.7%, when the number of clusters was five. In addition, the response time decreased from 0.0451 seconds to 0.0439 seconds when the number of clusters was three. This showed that the proposed data clustering algorithm decreased by 2.7% in response time as compared to the K-means data clustering. This study has shown that the proposed data clustering algorithm could be adapted by web search engine developers for more efficient web search result clustering.

BACKGROUND: Mobile devices have been impacting on human standard of living by providing timely and accurate information anywhere and anytime through wireless media in developing nations. Shortage of experts in medical fields is very obvious throughout the whole world but more pronounced in developing nations. OBJECTIVE: Thus, this study proposes a telemedicine platform for the vulnerable areas of developing nations. The vulnerable area are the interior with little or no medical facilities, hence the dwellers are very susceptible to sicknesses and diseases. METHODS: The framework uses mobile devices that can run LightWeight Agents (LWAs) to send consultation requests to a remote medical expert in urban city from the vulnerable interiors. The feedback is conveyed to the requester through the same medium. The system architecture which contained AgenRoller, LWAs, The front-end (mobile devices) and back-end (the medical server) is presented. The algorithm for the software component of the architecture (AgenRoller) is also presented. The system is modeled as M/M/1/c queuing system, and simulated using Simevents from MATLAB Simulink environment. RESULT: The simulation result presented show the average queue length, the number of entities in the queue and the number of entities departure from the system. These together present the rate of information processing in the system. CONCLUSION: A full scale development of this system with proper implementation will help extend the few medical facilities available in the urban cities in developing nations to the interiors thereby reducing the number of casualties in the vulnerable areas of the developing world especially in Sub Saharan Africa.

Spams are what users and developers should be aware of in all Internet-based communication tools (such as e-mail, websites, Social Networking Sites (SNS), instant messengers and so on). This is because spammers have not ceased from using these platforms to deceive and lure users into releasing vibrant and sensitive information (especially, financial details). This paper developed an architectural based technique for SPIM (Instant Message Spam or IM SPAM) detection using the classification method. The classification was done using the C4.5 classifier with a dataset of messages gotten from an instant messaging environment. The dataset served as the input to the classification algorithm method which was able to distinguish spam from non-spam messages. This classification method was depicted in a tree form to show its usefulness. The results show that its precision, recall and accuracy rate satisfied standard recommendation with a commendable error rate. The proposed technique will find implication in the reduction of the number of Internet users.Keywords: Social Networking sites, spammers, Instant message spam, C4.5 Classifiers, e-mails.