Forensic handwriting identification plays a pivotal role in criminal investigations and legal proceedings, requiring accurate and reliable techniques for analyzing handwriting samples. Convolutional Neural Networks (CNNs) are effective at feature extraction and classification for handwriting analysis, but their performance is highly sensitive to the choice of architectural hyperparameters, and manually tuned CNNs are prone to overfitting on forensic handwriting data. This study integrates the Chicken Swarm Optimization (CSO) algorithm with a CNN classifier (CSO-CNN) to automatically search the number of convolutional layers, the number and size of filters, and the batch size that minimize overfitting and maximize recognition accuracy. Handwritten samples were collected from staff and students of Ajayi Crowther University, Oyo, Nigeria, and organized into three variants (Original, Disguised, and Forged) with each variant split into training and test sets (72% and 28%, respectively) using a random-sampling hold-out procedure. CSO searched 30 candidate CNN configurations over the number of convolutional layers (1–3), the number of filters, the filter size, and the batch size, using recognition rate as the fitness function; the best-performing configuration used a single convolutional layer with 128 filters of size 3×3 and a batch size of 256. The resulting CSO-CNN model was compared against a standard, non-optimized CNN using accuracy, sensitivity, specificity, precision, false positive rate, and recognition time. CSO-CNN outperformed the standard CNN on every dataset variant: on the original dataset, accuracy improved from 82.14% to 86.43%; on the Disguised dataset, from 85.00% to 87.86%; and on the Forged dataset, from 83.57% to 89.29%, while recognition time was also reduced in all three cases. These results indicate that CSO-based hyperparameter search can meaningfully improve CNN-based forensic handwriting identification without requiring architectural redesign. As the evaluation reported here is based on a single hold-out split, repeated cross-validation and testing on independently collected datasets are recommended as a next step to confirm the statistical robustness and generalizability of these gains.
Primary healthcare systems, particularly in low and middle income countries, are caused by insufficient data, constraints on resources, and shifts in the demand for services. The focus of this paper is on a sys- tematic literature review to understand the impact of Generative Adversarial Networks (GANs) in the area of budget and planning in primary healthcare systems. Peer-reviewed literature was used for the analysis, and 35 studies published in the period of 2017 to 2025 were selected and reviewed based on the inclusion criteria of relevance to the area of forecasting, simulation, and decision support in healthcare data. Some of the key variations of GANs that are discussed in the paper involve the use of Conditional GANs, Condi- tional Tabular GANs (CTGANs) and other recurrent or temporal based GANs, which are commonly used for imaging, and electronic health record and time series data in healthcare. The generated findings suggest that, in resource-limited healthcare contexts, the use of GAN- based synthetic data generation improves the flexibility and consistency of predictive models, facilitates resource planning through scenario-based simu- lation, and allows for the controlled sharing of data across collaborating institutions. However, the review findings point to several gaps that include the absence of operational use in primary healthcare budgeting, difficulties in interpretability and governance of models, and no agreed-upon criteria for evaluating the trade-off between the reliability of synthetic data and associated privacy concerns. Therefore, the review demonstrates the possibilities and constraints of GAN-based methods and delineates paths for subsequent research to foster the integration of these methods into primary healthcare budgeting processes.
Software defect density prediction is essential for improving software quality and reliability; however, ac- curate prediction is challenging due to feature redundancy and imbalanced regression data. To address these issues, this study proposes a hybrid framework integrating Recursive Feature Elimination with Cross- Validation (RFECV), GridSearchCV, and Firefly Optimization (FFO) for feature selection and hyperpa- rameter optimization, along with SMOGN for imbalance handling. The framework is evaluated on six datasets using Gradient Boosting, Random Forest, Bagging, AdaBoost, Voting, and XGBoost regression models. Performance is assessed using MSE, MAE, RMSE, and SERA metrics. Experimental results show significant improvement, with RMSE reduced from 7.96 to 0.10 on the healthcare dataset and from 2.06 to 0.056 on the ANT 1.5 dataset. Comparisons with PSO, GWO, and GA further demonstrate superior accuracy, stability, and convergence behavior of FFO. Statistical validation using the Wilcoxon signed- rank test confirms the effectiveness and robustness of the proposed framework for software defect density prediction.
Artificial intelligence-based assistants have been developing at an incredible pace in recent years, facilitating fast information retrieval, automating repetitive tasks, and improving operational efficiency. In the field of software engineering, these tools now assist in activities such as testing, code generation, and documentation. Despite the increasing recognition of artificial intelligence (AI) as a valuable element of software development, its overall contributions remain insufficiently characterized. This research conducts a bibliometric analysis of AI-based assistants in the context of software development, identifying key publications, leading authors and organizations, and underexplored areas where these tools have a significant impact. As a result of a literature search conducted using the Scopus database, 83 papers (2014–2025) were analyzed using the Bibliometrix R-package for bibliometric evaluation. The collected documents revealed a sustained annual growth rate of 24.14%, with a peak in 2024 (31 papers), reflecting the surge in generative AI and LLM-based tools. The most cited article received 118 citations (39.33 citations per year), highlighting a strong impact in recent contributions. The United States led in publications with 26.1%, while Europe had the highest citation impact with 73 citations. Software design is the dominant theme with 41 papers and 19% of occurrences, while keyword trends focus on language models (LLM) with 19 papers, chatbots with 18 papers, and LLMs with 11 papers, confirming a clear shift toward LLM-centered research in software engineering. It provides insights into future research directions and opportunities for AI-driven innovation in software development .
