Ishaya Gambo

Introduction: Atopic dermatitis (AD) presents significant diagnostic challenges in darker skin types due to the masking of classical signs like erythema by pigmentation and the lack of diagnostic criteria tailored for this population. The use of machine learning (ML) as a noninvasive diagnostic tool in this context remains underexplored. Objectives: This study aimed to develop and evaluate the accuracy of an ML model in diagnosing AD in darker skin types and distinguishing it from healthy skin. Methods: A cross-sectional study using images of AD cases in dark-skinned patients was conducted. An EfficientNet-based convolutional neural network (CNN) model was trained on 1403 labelled images (687 atopic and 716 healthy skin images). Performance metrics including accuracy, precision, recall, and F1 score were assessed. Ethical approval was obtained, and datasets were anonymized for confidentiality. Results: The dataset comprised images from 50 AD patients (age range 1–93 years, median 18.0, IQR: 7.0–32.0) and 60 healthy volunteers (age range 18–50 years, median 27.5, IQR: 23.0–33.5), with a slight female predominance in both groups (56.0% and 53.3%, respectively). The model achieved an accuracy of 91.1%, precision of 83.7%, recall of 97.0%, and an F1 score of 89.9%, correctly classifying 103/106 AD images and 129/149 healthy skin images. External application programming interfaces (APIs) were integrated into the design to further enhance accuracy and generalizability. Conclusion: This study highlights the potential of ML as a diagnostic tool for AD in darker skin types. Further optimization and large-scale validation could enhance diagnostic accuracy, improving clinical outcomes for populations with darker skin. Iterative refinement and user feedback will ensure its continued efficacy and reliability in clinical practice.

The ever-evolving landscape of digital payments demands continuous innovation and self-improvement. This study addresses this imperative by simulating a model for payment routing, a crucial aspect of the digital payment ecosystem. To achieve this, industry professionals were interviewed to inform the approach, emphasizing data randomization for effective data collection. Using Python, a randomized dataset is created and three Reinforcement Learning (RL) algorithms are implemented and evaluated: Epsilon Greedy, Upper Confidence Bound (UCB), and Thompson Sampling. The paper adopts the Multi-Armed Bandit (MAB) framework to model payment routing as a resource allocation problem, offering a computational approach to real-world resource allocation dilemmas. Through simulation, we eliminate real-time transaction costs, allowing us to focus on algorithmic approaches without implications for customers, businesses, or payment providers. Among the RL algorithms studied, UCB emerges as the most effective in addressing this Multi-Armed Bandit problem, corroborating findings from prior research. This study suggests not only the potential of modeling real-world problems as MAB but also the superior performance of the UCB algorithm in solving RL problems. The paper underscores the need for increased focus on non-consumer-facing aspects of the financial services industry, emphasizing cross-disciplinary research to create infrastructure and software solutions. Researchers can extend this study by exploring MAB algorithms in various domains with options for system choices. The simulation-based approach offers a cost-effective means of testing system performance and hypotheses across a spectrum of industries, fostering innovation and progress.

Mobile app developers struggle to prioritize updates by identifying feature requests within user reviews. While machine learning models can assist, their complexity often hinders transparency and trust. This paper presents an explainable Artificial Intelligence (AI) approach that combines advanced explanation techniques with engaging visualizations to address this issue. Our system integrates a bidirectional Long Short-Term Memory (BiLSTM) model with attention mechanisms, enhanced by Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP). We evaluate this approach on a diverse dataset of 150,000 app reviews, achieving an F1 score of 0.82 and 89% accuracy, significantly outperforming baseline Support Vector Machine (F1: 0.66) and Convolutional Neural Network (CNN) (F1: 0.72) models. Our empirical user studies with developers demonstrate that our explainable approach improves trust (27%) when explanations are provided and correct interpretation (73%). The system’s interactive visualizations allowed developers to validate predictions, with over 80% overlap between model-highlighted phrases and human annotations for feature requests. These findings highlight the importance of integrating explainable AI into real-world software engineering workflows. The paper’s results and future directions provide a promising approach for feature request detection in app reviews to create more transparent, trustworthy, and effective AI systems.

Low-back pain (LBP) is a complex health problem requiring accurate diagnosis and effective treatment. However, the current decision support system (DSS) for LBP only considers the patient’s pain intensity and treatment suitability, which may not lead to optimal outcomes. This paper proposes a novel DSS that combines machine learning (ML) and expert input to classify LBP types and provide more reliable and personalized recommendations. We used an open-source dataset to train and test various ML models, including an ensemble model that combines multiple classifiers. We also performed data analysis and feature extraction to enhance the model’s predictive power. We developed a prototype tool to demonstrate the model’s performance and usability. Our results show that the ensemble model achieved the highest accuracy of 92.02%, followed by random forest (RF) (91.01%), multilayer perceptron (MP) (91.01%), and support vector machine (SVM) (87.88%). Our findings suggest that ML can help LBP specialists diagnose and treat LBP more effectively by learning from historical data and predicting LBP categories. Our DSS can potentially improve the quality of life for LBP patients and reduce the burden on the healthcare system.