Performance Evaluation of Selected Chaos-Enhanced Chicken Swarm Optimization for Handwritten Document Identification

Handwritten document identification (HDI) remains a challenging pattern recognition task due to high intra-writer variability and inter-writer similarity, which complicates accurate writer attribution. Traditional feature selection methods often fail to identify optimal feature subsets, leading to suboptimal classification performance and increased computational overhead. This research addresses these limitations by integrating chaotic maps into the Chicken Swarm Optimization (CSO) algorithm to enhance population diversity, prevent premature convergence, and improve feature selection capability. The primary aim is to evaluate the performance of three chaos-enhanced CSO variants (Chebyshev+CSO, Lorenz+CSO, and Henon+CSO) for HDI using the Institut für Informatik und Angewandte Mathematik (IAM) dataset. The methodology involves designing and implementing three chaos-enhanced CSO algorithms where chaotic sequences replace random number generators for population initialization, parameter adaptation, and position updates. The three chaotic maps are integrated into the CSO framework. The IAM handwritten dataset containing 500 samples with 9 features (stroke width, slant angle, character spacing, texture features, HOG features, and CNN features) across three writers is preprocessed. The models (Chebyshev+CSO, Lorenz+CSO, and Henon+CSO) were implemented in python, evaluated and compared based on accuracy, precision, recall, F1-score, and computational time. Chebyshev+CSO achieved 92.15% accuracy, 92.03% precision, 92.05 recall, 92.09% F1-score, converging speed 42 iterations, and 48.91s computational time. Lorenz+CSO attained 90.78% accuracy, 90.65% precision, 90.78 recall, 90.71% F1-score, converging speed 54 iterations and 52.67s computational time. While Henon+CSO achieved 89.23% accuracy, 89.01% precision, 89.23 recall, 89.12% F1-score, converging rate 67 iterations, and 45.23s computational time. The research establishes that chaos enhancement significantly improves CSO performance, with Chebyshev+CSO recommended for high-accuracy forensic applications, Henon+CSO for real-time systems, and Lorenz+CSO for complex exploration tasks. Major contributions are the comprehensive comparative analysis of chaotic maps, and identification of optimal feature subsets for writer identification.