Computer Science

Permanent URI for this collectionhttps://dspace.summituniversity.edu.ng/handle/123456789/22

Browse

Now showing 1 - 19 of 19

A COMPARATIVE STUDY OF LOGISTIC REGRESSION AND XGBOOST FO A COMPARATIVE STUDY OF LOGISTIC REGRESSION AND XGBOOST FOR CREDIT CARD FRAUD DETECTION R CREDIT CARD FRAUD DETECTION
(2024) KESHINRO OLUSHOLA SAMUEL
Since the dawn of recorded human history, fraud has entailed a variety of dishonest behaviors that vary greatly in their forms and tactics. Almost every purchase made today is done so online. Online transactions are completed using an easy-to-use, multi-party, straightforward approach that does not require the usage of cash. The annual loss resulting from fraudulent credit card transactions is in the billions. According to the 10th annual study on online fraud, between 2006 and 2008, 1.4% of online payments resulted in lost money; nevertheless, the real percentage of lost revenue increased as online sales increased. The annual loss resulting from fraudulent credit card transactions is in the billions. The 10th annual study on online fraud states that while 1.4% of online payments resulted in lost money between 2006 and 2008, the real percentage of lost revenue increased as online sales increased. The present dataset for this study was collected in September 2013 through credit card transactions made through Kaggle by cardholders throughout Europe. In this study, a model built with Logistic Regression techniques was compared to the XGBoost model that was based on standard evaluation criteria.
A FACE DETECTION AND RECOGNITION-BASED ONLINE ATTENDANCE SYSTEM USING COMPUTER VISION
(2023) NAFISAH, ADEDAYO SULAIMAN.
Face recognition has drawn a lot of attention recently and is a crucial issue in many applications, including access control, security systems, and credit card verification and identification of criminals. This study suggests three primary subsystems, including autonomous door access control, face detection, and face recognition. By adapting the principal component analysis (PCA) approach to the fast based principal component analysis (FBPCA) approach, the face identification and detection process is achieved. The captured image is recognized using a web camera and compared with the image in the database. To achieve the goal of identification, image processing and recognition are applied to the actual image modification and transformation. This project focuses on the design and of a facial recognition attendance system using computer vision technique. The system aims to automate the process of identifying and verifying individuals in an organization based on their facial features. By leveraging advanced algorithms, image processing techniques, and deep learning models, the system achieves accurate and real time facial recognition. The project involves data collection, pre-processing, feature extraction, and system integration to develop a comprehensive facial recognition solution
A HYBRID LINEAR PROGRAMMING MODEL AND GENETIC ALGORITHM APPROACH FOR RESOURCES ALLOCATION IN DISASTER RESPONSE
(2023) SHOBAYO, SULAIMON
Efficient allocation of resources such as emergency personnel and equipment plays a major role in disaster scenarios by minimizing the response time. Limited resources concerns using resources as productively as possible. This research focuses on novel development of a hybrid linear programming model and genetic algorithm approach for resource allocation in accident disaster response. The objective is to optimize the total response-time for allocation of resources to affected areas and populations and enhance the efficiency and effectiveness of disaster response operations. This model combines the strengths of integer linear programming, which provides a systematic framework for the minimization of the total response time, and the Genetic Algorithms, which handle the case of complex and dynamic problem spaces. The GA leverages on the model formulated in the ILP trade-off to handle the complex based spaces which are utilized to search for near-optimal solutions within the problem space. The model considers various factors such as distance between resources and number of affected areas, capacity of affected area, and resource capacities. The objective function minimizes the response time by optimizing the distance between the resources and number of affected areas, while the constraints are, resources allocation, and capacity area. The result shows an improved and efficient outcome to response operations in minimizing response time to reaching affected area and maximizing coverage area with available resources. The outcomes enables decision-makers to make informed and optimized choices during critical situations by improving overall response outcome.
AN IMPROVED ARTIFICIAL INTELLIGENCE MODEL FOR CYBER SECURITY INCIDENT RESPONSE AND RECOVERY SYSTEM
(2024) ASIMI OLALEKAN IDRIS
Cyber security incident response and recovery systems are currently facing a number of challenges that are different from the fast growth of advanced cyber threats to the complexity faced in coordinating an effective response across the various technological environments. Several techniques have been developed, but there are problems in the detection and mitigation of emerging threats in real-time, thus, organizations are at risk of data breaches, financial losses, and reputational damage. This study presents an improved Artificial Intelligence model that assists in effective incidence response and recovery from previously known and unknown threats. The bagging ensemble approach is adopted using Naïve Bayes, Decision Tree, Support Vector Machine and Neural Network as base classifiers to form the model. In the experiment, the dataset used has a total of 22544 instances and 42 attributes. The result gives 98.69% accuracy with ROC and PRC Area both 0.999. The Recall and F-Measure are both 0.987.
ANALYZING CUSTOMER SUPPORT LOGS FOR ANOMALY DETECTION USING DEEP LEARNING
(2024) ISSA, SHAMSUDEEN ENIOLA
This project investigates the application of deep learning techniques for anomaly detection in customer support logs, a critical resource for insurance companies. The increasing volume of customer interactions generates vast amounts of data, making it challenging to manually identify irregularities that may indicate fraud or data entry errors. By leveraging advanced data-driven methods, we developed a robust deep learning model capable of analysing historical support logs to detect anomalies with high accuracy. The study employs various neural network architectures, including convolutional and recurrent networks, to capture complex patterns within the data. We also explore feature engineering techniques to enhance model performance and ensure the reliability of the detection process. The results demonstrate significant improvements in identifying anomalies compared to traditional methods, underscoring the potential of deep learning in automating and streamlining customer support operations. Furthermore, the project emphasizes the importance of enhancing data quality, implementing real-time detection systems, and automating alerts to improve overall customer support efficiency. By addressing these challenges, our findings contribute to the development of more effective strategies for managing customer interactions and safeguarding against potential fraud, ultimately leading to better service delivery and customer satisfaction.
APPLICATION OF ANIMATION TO TEACHING (A CASE STUDY OF SUMMIT UNIVERSITY, OFFA, KWARA STATE)
(2022) RAJI MUBARAK AKINOLA
Computer-mediated teaching & learning in higher institutions classroom is much relevant in recent times. Effective, efficient and knowledge-driven classroom teaching should not be limited to mere chalk, board and teacher relationship. Animation made Teaching and learning innovative, enjoyable, impacting and understandable. Whiteboard animation is a kind of animation that enable teaching with broad explanation, making it suitable for the creation of the videos in this project. Whiteboard animated videos were designed and created for selected course in the department of computer science. The videos were uploaded on a Youtube channel to aid easy access. The video’s texts were implemented using Adobe After Effect and the voiceover were recorded with Adobe Audition. Animated videos will enhance teaching in a wide range if it is adopted in every educational level.
ARTIFICIAL INTELLIGENCE BASED IRIS SEGMENTATION
(2023) SAHEED GANIYAT YAHAYA
The iris, a colored component of the eye surrounding the pupil, possesses a unique and distinctive pattern for each individual. Iris segmentation plays a pivotal role in the field of biometric. It serves as a critical stage in iris recognition systems, separating the iris area from other parts of the eye image, thereby enhancing the effectiveness of subsequent stages. In this paper, we propose a fusion-based iris segmentation technique that combines Thresholding and Limbic Pupil Boundary (LPB) methods. We evaluate the robustness of our approach using the CASIA and Ayush datasets. Specifically, we perform a comparative analysis on the Ayush dataset, comprising 650 eye images. The results of the evaluation demonstrate that the proposed approach accurately segments the iris from the eye image, laying the foundation for a reliable recognition system
CRIME ANALYSIS AND PREDICTION USING DATA MINING TECHNIQUES
(2023) IBRAHIM SODIQ OLANIYI
CRIME ANALYSIS HAS EVOLVED SIGNIFICANTLY WITH THE INTEGRATION OF DATA MINING TECHNIQUES, ENABLING LAW ENFORCEMENT AGENCIES TO UNCOVER HIDDEN PARTTERNS AND PREDICT POTENTIAL CRIMINAL ACTIVITIES. THIS STUDY EXPLORES THE APPLICATION OF DATA MINING ALGORITHMS TO ANALYZE CRIME DATA AND FORECAST FUTURE TRENDS.
DESIGN AND IMPLEMENTATION OF AN ANDROID-BASED CUMULATIVE GRADE POINT AVERAGE CALCULATOR
(2022) ABDULWASIU JAMIU OLASUNKANMI
UI/UX refers to the practice of designing digital products with a user-first approach. An android mobile device enables open-source applications embedding UI/UX interface, which enable its high percentage of usage in the student populace. It is being discovered that poor performance of students in assessments can be linked to limited study and concentration which has been traced to the common lack of drive. The GPA calculator App is developed to drive the student to study & prepare hard for the assessment and even predict the grade. This project was developed using java programming language and React JS. The developed GPA Calculator has been hosted on Netlify and published on Google Play Store. The GPA calculator is effective to drive the student to perform well in pursuing good grades.
DESIGN AND IMPLEMENTATION OF ENGLISH TO YORUBA LANGUAGE TRANSLATION SYSTEM
(2024) FATHIAT ABIMBOLA ODUTAYO
Research in machine translation (MT) is advancing in recent times, however, concentration has been on languages with high number of speakers and abundant digital resources. Unfortunately, some languages, like Yoruba, are often left out. This study focuses on the development of an English to Yoruba machine translation system, addressing the need for a translation system to allow for ease of communication between Yoruba speakers who doesn’t understand the English language. A substantial parallel corpus, primarily sourced from the Menyo-20k dataset, was employed to ensure diverse topic coverage and robust evaluation. The methodology involved data collection, preprocessing, model development, and evaluation using the BLEU metric. The resulting system demonstrated a BLEU score of 11.9, indicating foundational translation capabilities while highlighting the need for further refinement in idiomatic expressions. This work not only contributes to the preservation and promotion of the Yoruba language but also serves as a framework for future research in low-resource language translation systems, emphasizing the importance of cultural sensitivity in technological advancement
DEVELOPMENT OF AI ENHANCED DARK WEB DETECTION SYSTEM WITH QUANTUM CRYPTOGRAPHY
(2024) MUKHTAR OLATUNDE DUNMOYE
Quantum computing combined with machine learning provides an environment that allows implementing better security measures within that realm. This methodology provides cutting-edge Quantum Machine Learning algorithms that turn out to be useful in information security measures. Quantum computing is a method of computing based on the principles of quantum mechanics, which allows quite incredibly fast computation, way beyond the standard forms previously available, and new solutions in encryption for secure communications and data storage. In information security, quantum machine learning algorithms are trained using quantum computing, where the quantum computer is used in processing enormous datasets to discover trends and flag anomalies that cannot be explained. The paper surveys some of the QML methods, including quantum support vector machines, quantum neural networks, and quantum clustering algorithms, and how they perform better than traditional methods for intrusion detection, malware classification, and mode detection. This paper also addresses issues and prospects related to quantum machine learning, such as quantum cryptography, which involves robust quantum-resistant algorithms and encryption systems in the face of quantum. It finally pointed out quantum machine learning as a new avenue for cyber security in cyberspace where illegal cyber activities are rapidly evolving and quantum technologies are emerging
DEVELOPMENT OF ARTIFICIAL INTELLIGENT-BASED METAL DETECTION IN DRINKING WATER
(2024) AYINLA, RIDWAN OLAYIWOLA
DEVELOPMENT OF FACIAL RECOGNITION-BASED TODDLER’S EMOTION PREDICTION SYSTEM
(2023) BASHIRU, BASIT AYOMIDE
Toddler’s emotion is a set of expressions; facial or verbal ranging differently in toddlers, these expressions are usually determined by the environment. However, inability to detect and predict change in toddler’s emotion lead to late intervention in toddler’s development resulting to a negative impact on the mental and social development. To proffer a long-term solution to this problem, the study developed a facial recognition-based toddler’s emotion prediction system. The system’s model is developed using random forest algorithm and trained with the features extracted from the Toddler’s happy and sad facial dataset of 2168 image sample size. Feature extraction of the images is done using Mediapipe machine learning algorithm. The model was integrated into a designed user interface for ease of use. The interface captures the toddler’s face image and makes emotion predictions based on happy or sad. In conclusion, the developed facial recognition-based toddler’s emotion prediction model performs excellently with an accuracy score of 84%. With the help of the recognition rate of the developed system, the study is to predict toddler’s emotion based on two classification which can either be happy or sad at real-time which will aid parent or caregiver on how to treat the toddler based on emotion detected.
DEVELOPMENT OF FACIAL RECOGNITION-BASED TODDLER’S EMOTION PREDICTION SYSTEM
(2023) BASHIRU, BASIT AYOMIDE.
Toddler’s emotion is a set of expressions; facial or verbal ranging differently in toddlers, these expressions are usually determined by the environment. However, inability to detect and predict change in toddler’s emotion lead to late intervention in toddler’s development resulting to a negative impact on the mental and social development. To proffer a long-term solution to this problem, the study developed a facial recognition-based toddler’s emotion prediction system. The system’s model is developed using random forest algorithm and trained with the features extracted from the Toddler’s happy and sad facial dataset of 2168 image sample size. Feature extraction of the images is done using Mediapipe machine learning algorithm. The model was integrated into a designed user interface for ease of use. The interface captures the toddler’s face image and makes emotion predictions based on happy or sad. In conclusion, the developed facial recognition-based toddler’s emotion prediction model performs excellently with an accuracy score of 84%. With the help of the recognition rate of the developed system, the study is to predict toddler’s emotion based on two classification which can either be happy or sad at real-time which will aid parent or caregiver on how to treat the toddler based on emotion detected.
DEVELOPMENT OF QUANTUM SECURE AI CHATBOT
(2024) HUSSEIN SOLIHAT BIDEMI
Chabot continuously increase customer services through immediate 24/7 support, handling simple questions, and releasing human agents to more complicated questions, which ultimately improve the productivity and satisfaction of users. However, there is an increasing concern about intruders getting access to users’ personal details through chatbot. To address this potential privacy concerns due to the handling of sensitive user data. This study presents a quantum safe intelligent chatbot. The design of a quantum safe intelligent chat bot signifies intersection area between quantum computing technology, Artificial Intelligence and Cyber security. This study adopts quantum key distribution principle in combination with AES algorithm to ensure secure and reliable communication while using chatbots.
ENERGY OPTIMIZATION IN SMART GRIDS WITH DEEP REINFORCEMENT LEARNING
(2024) AJIKOBI MUBARAK DAMILOLA
The escalating complexity, uncertainty, and data volumes in energy systems have rendered conventional methods ineffective in addressing decision-making and control challenges. As a result, data-driven approaches have become a crucial focus area. Deep reinforcement learning (DRL) represents a significant breakthrough in data-driven technology, earning its reputation as a true form of artificial intelligence (AI). By combining the capabilities of deep learning (DL) and reinforcement learning (RL), DRL gives rise to a robust and adaptive approach that excels in complex decision-making and control scenarios. With its successful applications in various domains, DRL has been increasingly applied to optimize energy systems, including energy management, demand response, smart grids, and operational control. This paper provides a thorough review of DRL's fundamental principles, models, and algorithm, followed by an in-depth exploration of its applications in energy optimization. Furthermore, the paper discusses recent breakthroughs in DRL, including its integration with traditional methods, and examines the opportunities and challenges of its applications in the energy sector
INTEGRATION OF U-NET AND MASK R-CNN APPROACH FOR AUTOMATED CLASSIFICATION OF HISTOPATHOLOGY IMAGES OF PROSTATE CANCER.
(2024) ADEYEMI KHALILULLAHI ADEBAYO
This study proposes a novel approach for automated classification of histopathology images of prostate cancer by integrating U-Net and Mask R-CNN models. The U-Net model is designed to perform segmentation, localizing regions indicative of prostate cancer, while the Mask R-CNN model is optimized for object detection, enhancing classification precision. Our model implementation leverages modified TensorFlow and Mask R-CNN configurations, utilizing custom dataset generation and preprocessing pipelines to handle labeled prostate histopathology images. The dataset is split into training and testing sets, with the U-Net model trained for segmentation tasks, supported by a data generator class for efficient batch processing. After training, the model’s performance is evaluated using metrics such as accuracy, F1 score, recall, and AUC. The initial results demonstrate promising capabilities in accurately segmenting and classifying cancerous regions in histopathology images, indicating the potential for improving diagnostic accuracy in prostate cancer.
MATHEMATICAL MODELING OF ROAD TRAFFIC FLOW IN URBAN AREAS WITH NEURAL NETWORK
(2024) BABALOLA, BABALOLA
study of traffic flow in urban areas is of paramount importance due to its significant impact on transportation efficiency, environmental sustainability, and overall quality of life. Mathematical modeling serves as a powerful tool to analyze and understand the intricate behaviors exhibited by vehicular traffic within urban settings. This study explores the development and application of mathematical models to characterize the complex interactions among vehicles, pedestrians, infrastructure, and environmental factors in urban traffic systems. The study uses ANN based model for detecting the traffic flow in the urban. And the model is trained on dataset from Kaggle for the ANN algorithms with the mathematical liner equations. Additionally, the study delves into the incorporation of factors such as traffic signals, road geometries, driver psychology, and emerging technologies like connected and autonomous vehicles into these models. Through the synthesis of empirical data, advanced simulation techniques, and theoretical analysis, mathematical models offer valuable insights into traffic management strategies, congestion mitigation, urban planning, and the development of intelligent transportation systems for future cities
THE USE OF AI TO DETECT DROWSINESS IN DRIVERS
(2023) ADEYEMO, MUKTAR
Driver distraction and drowsiness are critical concerns leading to road accidents, causing significant harm in Malaysia and worldwide. This study addresses the pressing need for effective driver distraction detection systems to enhance road safety. We propose a comprehensive methodology comprising data preprocessing, data augmentation, and a novel Driver Drowsiness and Monitoring System. This system incorporates a lightweight model, EfficientNetB0, coupled with a Channel Attention (CA) mechanism, demonstrating superior performance. In the experimentation phase, we evaluate our model on two benchmark datasets: State Farm Distracted Driver Detection (SFD3) and AUC Distracted Driver (AUCD2). The results indicate that our model achieves remarkable accuracy, precision, recall, and F1-score, surpassing existing state-of-the-art models. Moreover, our model exhibits exceptional time efficiency, making it suitable for real-time applications and resource-constrained devices. This research contributes to mitigating distracted driving's adverse effects, ultimately reducing accidents and promoting safer driving practices. The proposed methodology and results showcase the potential for deploying such systems to enhance road safety and reduce road accidents.