Professor Celestine Iwendi

The creation of digital marketing has enabled companies to adopt personalized item recommendations for their customers. This process keeps them ahead of the competition. One of the techniques used in item recommendation is known as item-based recommendation system or item-item collaborative filtering. Presently, item recommendation is based completely on ratings like 1-5, which is not included in the comment section. In this context, users or customers express their feelings and thoughts about products or services. This paper proposes a machine learning model system where 0, 2, 4 are used to rate products. 0 is negative, 2 is neutral, 4 is positive. This will be in addition to the existing review system that takes care of the users' reviews and comments, without disrupting it. We have implemented this model by using Keras, Pandas and Sci-kit Learning libraries to run the internal work. The proposed approach improved prediction with 79% accuracy for Yelp datasets of businesses across 11 metropolitan areas in four countries, along with a mean absolute error (MAE) of 21%, precision at 79%, recall at 80% and F1-Score at 79%. Our model shows scalability advantage and how organizations can revolutionize their recommender systems to attract possible customers and increase patronage. Also, the proposed similarity algorithm was compared to conventional algorithms to estimate its performance and accuracy in terms of its root mean square error (RMSE), precision and recall. Results of this experiment indicate that the similarity recommendation algorithm performs better than the conventional algorithm and enhances recommendation accuracy.

Accurate stock price forecasting remains a challenging yet crucial task in the financial industry due to the non-linear relationships, noisy, and time-dependent nature of the market data. This study presents a deep learning approach known as long-short-term memory (LSTM) for predicting the closing prices of stock using historical data. The model is designed to capture the complex temporal dependencies inherent in stock market sequences, addressing the limitations of traditional statistical models such as ARIMA and linear regression. Using key key characteristics such as past closing prices, the LSTM model achieved high predictive performance with a Mean Squared Error (MSE) of 0.00036, a mean absolute error (MAE) of 0.0096, and a coefficient of determination (R²) of 0.9941, indicating strong generalization and accuracy. The results demonstrate the effectiveness of LSTM architectures in time series forecasting for financial applications. This research contributes to the development of robust and automated decision support tools for investors and sets a performance benchmark for future deep learning models in stock market prediction.

The WannaCry ransomware attack that happened in May 2017 represented a turning point for the modern cybersecurity landscape and, at the same time, spawned many lines of ethical debate related to discovering, using, and disclosing software vulnerabilities. This paper discusses ethical lessons from the WannaCry attack; it explores what this might mean for the respective roles and responsibilities of governments, technology companies, and cybersecurity professionals in managing zero-day vulnerabilities. It contemplates the broader implications for society as a whole of such decisions, and tensions between interests of national security and those of global cybersecurity. Ethical frameworks guiding future cybersecurity practices are proposed in the conclusion

Sentiment analysis is a subfield of natural language processing (NLP) that aims to determine the emotional tone and sentiment expressed in a given piece of text. It plays a crucial role in understanding the opinions, attitudes, and emotions of individuals towards various subjects, products, or events. With the rapid growth of online communication and social media, sentiment analysis has become increasingly important for businesses, governments, and researchers to gain valuable insights into public sentiment and make data-driven decisions. Deep Learning, a branch of machine learning, has shown remarkable success in various NLP tasks, including sentiment analysis. This research explores role of sentiment analysis in twitter data, application of deep learning techniques in sentiment analysis, focusing on recurrent neural networks (RNNs), convolutional neural networks (CNNs), and transformer-based models. In this paper sentiment analysis perform on a dataset of tweets related to the Pfizer vaccine. Here we have data related to vaccination tweets of 11021 users and sentiments of users. It processes text input, classifies feelings, and visualizes the findings using multiple natural language processing techniques and machine learning algorithms.

This paper presents a novel Intrusion Detection System (IDS) framework for securing Internet of Things (IoT) networks, leveraging advanced machine learning techniques. The proposed framework integrates Deep Neural Networks (DNNs) and Random Forest (RF) algorithms to enhance detection accuracy and robustness. Utilising the comprehensive CICIoT2023 dataset, the IDS model is rigorously trained and evaluated, demonstrating high efficacy in detecting and mitigating potential threats. However, the results also reveal shortcomings in detecting certain attack categories, such as command injection and SQL injection, indicating areas for further refinement. These findings contribute to the advancement of IoT security through the application of advanced machine learning techniques, while also highlighting the need for continued research to address identified shortcomings.

This paper presents a novel hybrid deep learning model that combines Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) for accurate age and gender classification from iris images. Using the GMBAMU-IRIS dataset, the model leverages CNNs for spatial feature extraction and RNNs for sequential pattern recognition. The proposed model achieved an age classification accuracy of 85.68% and a gender classification accuracy of 98.95%, outperforming traditional methods. These findings highlight the potential of hybrid models in enhancing biometric recognition systems. 1 Introduction Biometric recognition systems have emerged as critical components in various security and authentication applications, providing reliable and efficient means of identifying individuals based on their physiological and behavioral traits. Among the array of biometric modalities, the human iris stands out due to its unique patterns and remarkable stability throughout an individual's life. Iris recognition systems leverage the distinct textural patterns within the iris to achieve high levels of accuracy and reliability in personal identification and verification tasks [1][2]. Beyond the primary goal of identification, there is a growing interest in using iris images for soft biometric traits, such as age and gender classification. Accurate age and gender classification can significantly enhance the functionality of biometric systems by providing additional contextual information [3]. This additional layer of data can improve user experience , enable demographic-specific applications, and enhance the robust-ness of security systems by adding another factor of authentication [4].

Uterine cancer prediction accuracy is important in clinical decision-making because it improves the overall chances of patient recovery. Several machine learning models, such as Decision Tree, Random Forest, XGBoost Regressor, and Support Vector Regressor, were explored to determine which is more effective in predicting uterine cancer. Attributes such as mutation counts, diagnosis age, and MSI score, were used for the analysis. The different models were tested using the standard performance metrics such as the Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R2 Score. Random Forest showed the highest predictive performance with an R2 score of 0.655, followed by XGBoost regressor, which was relatively close to the R2 score of Random Forest. Support Vector Regressor performed very poorly as the R2 score was negative, implying that the model is not suitable for such prediction. Ensemble-based models, which include Random Forest and XGBoost Regressor, have proven to be more effective in handling medical prediction tasks, and this is because of their robustness and their ability when it comes to handle overfitting. Though model generalizability was affected due to small data size and the absence of hyperparameter tuning. The future work will focus on expanding the dataset, implementing hyperparameter tuning, integrating deep learning, and leveraging explainable AI (XAI). The research has provided valuable insight for clinicians who wish to use machine learning for uterine cancer prognosis.

This study presents a deep learning-based approach to predicting physiological health risks in athletes engaged in contact sports using wearable sensor data. Motivated by the need to detect early warning signs of collapse or severe fatigue, this study employs a Long Short-Term Memory (LSTM) neural network to analyse multivariate time-series data. Key physiological signals, including heart rate, body temperature, and motion, were extracted from the PAMAP2 dataset to train and validate the model. The LSTM demonstrated strong predictive performance, achieving an accuracy of 98.3% in identifying potentially dangerous physiological states. In addition to its high classification accuracy, the model effectively captured temporal dependencies in the data, underscoring its suitability for health risk prediction in dynamic, high-intensity sports environments. This study highlights the potential of wearable data and LSTM-based analysis in supporting proactive athlete health management and injury prevention.

Microwave radio links play a vital role in keeping mobile networks running, especially when it comes to backhaul-the part of the network that connects base stations to the core. Gradual failure in these links could disrupt services and cost providers a lot in both revenue and customer trust. In this study, we explore how machine learning can help predict such failures before they happen. Network performance data from a mobile network operator in Nigeria was collected, cleaned and used to achieve the purpose of the study. Four algorithms belonging to machine learning (ML) and deep learning (DL) were adopted and used for training the dataset and predicting link failures. Results show that the Long ShortTerm Memory (LSTM - a deep learning model effective for handling time-series data) performed best with prediction accuracy of 92%, distantly followed by others. These findings indicate that the LSTM is better in modelling temporal patterns in network behaviours. This study provides a practical framework for automating microwave link monitoring and maintenance, thereby reducing manual diagnostics, preventing outages, and improving service reliability. The proposed solution supports the integration of predictive intelligence into network operations, enhancing the quality of service and operational efficiency for telecom providers.

HALOGrid is an adaptive edge–cloud malware detection framework for IoT traffic. The approach couples a lightweight LSTM (residual paths, attention, drift-penalty regularization) for low-latency edge inference with a telemetry-driven tuner that performs real-time hyperparameter updates. The tuner employs Augmented Grid Search (AGS): a stage-wise coarse-to-fine exploration with stochastic perturbations, early-stopping of inferior candidates, validation-weighted corrections, and expectation-weighted deployment. A resynchronization controller blends edge and cloud states using divergence- and delay-aware gating; updates are secured via mTLS transport and signed artifacts with rollback. The pipeline integrates preprocessing, drift estimation over multi-metric streams, adaptive learning-rate/regularization adjustment, and A/B deployment safety. Evaluation on CICIoT2023 reports 98.74% accuracy, 1.21% false positive rate, and 12.8,ms mean inference latency on Jetson Nano; energy consumption averages 52.5,mJ/inference. Compared with SGM, HPAI, DFN, ODMS, MIHT, AIMO, IEMS, and DOFD, HALOGrid maintains higher detection fidelity with lower tuning overhead through AGS and secure edge–cloud refinement.