Abstract

Handwritten document identification (HDI) was an important task in forensic analysis, digital archiving, and automated recognition systems. However, variations in handwriting styles, noise, and document quality made accurate identification challenging. Although Chicken Swarm Optimization (CSO) had potential for solving complex optimization problems, it could suffer from premature convergence, poor population diversity, and reduced performance in complex search spaces. This study therefore addressed the need to evaluate the comparative performance of three chaos-enhanced CSO variants (Circle+CSO, Gauss+CSO, and Henon+CSO) for HDI using the Institute of Computer Science and Applied Mathematics (IAM) dataset. The methodology involves designing and implementing three chaos-enhanced CSO algorithms where chaotic sequences replace random number generators for population initialization, rooster update mechanisms, hen following behaviors, chick movements, and perturbation strategies. The IAM Handwriting Database containing 500 samples with 9 features (stroke width, slant angle, character spacing, texture 1 and 2 features, and CNN 1 and 2 features) across three writers is preprocessed. The models (Circle+CSO, Gauss+CSO, and Henon+CSO), were implemented in python, evaluated and compared based on accuracy, precision, recall, F1 score, converging rate and computational time. Henon+CSO achieved 93.12% accuracy, 90.21% precision, 90.34% recall, 90.27% F1 score, converging rate 41 iterations, and 2.45s computational time. Circle+CSO attained 90.34% accuracy, 90.21% precision, 90.34% recall, 90.27% F1 score, converging rate 58 iterations, and 2.38s computational time. While Gauss+CSO achieved 91.89% accuracy, 91.78% precision, 91.98% recall, 91.83 F1 score, converging rate 47 iterations, and 2.52s computational time. This research demonstrates that integrating chaotic maps into the CSO framework improves handwritten writer identification performance. The study contributes to knowledge by systematically comparing Circle, Gauss, and Henon chaotic maps and identifying their respective strengths for different applications. Henon+CSO is the most suitable for high-accuracy and fast-convergence applications, Circle+CSO for computationally time-critical applications, and Gauss+CSO for maintaining population diversity and preventing stagnation. The study also identifies effectivefeature subsets, providing a useful foundation for improved handwritten document identification.