Iraqi Journal of Data Science

Optical Coherence Tomography (OCT) greatly facilitates the diagnosis of retinal diseases. However, traditional models based on Convolutional Neural Networks (CNNs) suffer from challenges, most notably high computational cost, sensitivity to noise, and data imbalance. This study aims to compare three hybrid deep learning frameworks, all of which rely on feature extraction using a pre-trained CNN model and then selecting the most important features using intelligent swarm algorithms: the Dolphin Swarm Optimization (DSO), the Particle Swarm Optimization (PSO), and the Ant Swarm Optimization (ACO). The selected features were evaluated using four classifiers: SVM, random forest, XGBoost, and k-NN. Experiments were conducted on a standard dataset from the University of California, San Diego (UCSD) and a local dataset. The comparison results showed that the hybrid framework, which combines the dolphin swarm algorithm and SVM, outperformed the other combinations, achieving a classification accuracy of 93% on local data and 95% on standard data, while also outperforming them in terms of accuracy and computational efficiency.

The online payment system is at high risk due to the increasing rates of credit card theft. The primary objective is to identify cases of credit card theft by analysing the purchase history of cardholders and categorising them accordingly. These include an increase in the slope of logistics, steep slope, and scattered woodlands. The proposed model utilises tools such as logistic regression and random forest as machine learning techniques. Additionally, a set of preprocessing techniques is employed, including data balancing using SMOTE. After being trained on a large dataset of credit card transactions, the model is used to detect trends and anomalies that may indicate fraudulent activity, taking into account factors such as transaction amount, location, and time of day. We have used artificial minority oversampling to put the data set into proper perspective. The two algorithms were applied, yielding 97.34% accuracy for Logistic Regression and 99.99% accuracy for Random Forest. The accuracy metric is used for performance evaluation. The results indicate a promising performance that can enhance credit card security, potentially helping to reduce financial losses to victims of fraud.

The uncontrollable increase in the number of connected devices and smart homes has posed a significant security threat, particularly to systems such as smart locks, HVAC systems, and networking protocols like Zigbee and Z-Wave. The issue discussed in this paper is the issue of poor monitoring and threat detection in smart homes. Our proposal and implementation involve a prototype security module that can be used to monitor a smart home network, detecting any suspicious activity and notifying users accordingly. To measure the effect on system resources of the prototype and to observe its capability to detect malicious behavior, the prototype was implemented and tested on a smart hub platform. The experimental findings suggest that the module has the lowest performance overhead whilst having a high detection rate. Such results suggest that small-scale security surveillance can be utilized to enhance the robustness of smart home systems against cyberattacks and increase the security of individuals and the community.

In the digital era, substantial amounts of textual information are embedded in images, especially across news outlets, social platforms, and scanned documents. This presents a significant technical challenge: efficiently extracting and summarizing text from images in an automated way that preserves context and meaning. Traditional text summarization techniques are not directly applicable to image-based content because they depend on pre-structured input text. In this paper, we propose a framework that integrates Optical Character Recognition (OCR) and advanced Natural Language Processing (NLP) models to address this challenge. The proposed method implements OCR to extract raw text from images, followed by deep learning-based summarization using models such as LSTM, Bi-LSTM, BERT and T5. These models are trained on large-scale news datasets to enhance their ability to generate coherent summaries from unstructured text. To ensure accessibility and practical usability, our framework is deployed via an interactive web-based interface that allows end-users to upload images and receive concise summaries in real time. Experimental evaluation demonstrates the efficacy of the proposed approach, particularly with transformer-based models, in delivering high-quality summarization from visual text sources

Testability, as an essential property, plays a crucial role in software quality and testing techniques. In this work, we are conducting a comparative study of TDD and TsDD. Whereas TDD focuses on generating unit tests before implementing the code, TsDD aims to evaluate the testability of code before running static analysis techniques and changing the design before carrying out test execution. This study is quantitative and focuses on collecting and analysing numerical data, using code coverage (CC) and testing timeframe as two important criteria for evaluating software development and testing frameworks. The impact of this approach on software quality is measured using advanced data testing instruments, monitoring code coverage, and testing time trade-offs across three open-source software projects. Results show that code coverage is significantly improved and that the software becomes more testable, with TseDD outperforming TDDe by 14.20% on testability and 12.54% on code coverage. This underscores the significance of this technique as a successful approach to improving software quality and streamlining development processes.

Artificial Intelligence (AI) has made advancements, among other things, OpenAI created the sophisticated model ChatGPT. Conversational, ChatGPT supports natural interactions, providing human-like responses to queries across myriad topics. But it is not infallible, and the degree of accuracy also depends on the complexity of the queries, the context, and how often the prompts are repeated. This work thus proposes a new model, the Precision Answer Comparison and Evaluation Model (PACEM), to systematically address these types of questions and assess ChatGPT's performance. PACEM assesses the correctness and coherence of ChatGPT's answers across numerous fields, including literature, history, law, ethics, and sports. By providing these analyses and comparisons, PACEM goes on record with a detailed understanding of what ChatGPT does well and poorly as a source of reliable information. On top of that, it includes an assessment of response time, considering ChatGPT's speed in producing answers in relation to real or expected ones. The findings show that ChatGPT's answers are usually substantially accurate and often of superior quality compared to those written by the user and other alternatives. Response time generally increases with the complexity or length of the answer. Finally, the study reviews notable takeaways from PACEM's deployment and offers suggestions for future research to address the evolving challenges in AI-driven response assessment.