HCVSOA: hybrid corona virus–snake optimization for hyperparameter-tuned kinematic handwriting analysis in Parkinson’s disease detection

Abstract Working with machine learning models is challenging due to various complex factors such as high-dimensional temporal dependencies, control complexity and poor subject-independent performance. Researchers have identified that the conventional optimization strategies often fail to explore these complex factors fully which limits model robustness and subject-independent performance. The study introduces hybrid heuristic optimization strategy namely, Hybrid Corona Virus-Snake Optimization Algorithm (HCVSOA) which combines the exploratory strengths of Corona Virus Optimization Algorithm (CVOA) with the adaptive exploitation of the Snake Optimization Algorithm (SOA). The HCVSOA framework is capable of tuning model parameters for machine learning models which facilitates more effective temporal feature extraction and structured feature representation for strengthening accurate class-based classification. Experiments are conducted on both handwriting dynamics datasets with meander and spiral patterns. The HCVSOA + Convolutional 1D Bidirectional Gated Recurrent Unit framework demonstrates superior performance over other considered optimization strategies which records an accuracy of 96.25% on meander New HandPD signals, 97.86% on spiral New HandPD signals and 99.903% on spiral UCI dataset. The observed improvements in performance metrics confirm a well-balanced trade-off between sensitivity and specificity. Further, Area under the Curve and radar analyses confirm enhanced discriminative capability which highlights the consistent advantage across architectures and datasets for hybrid optimization techniques. The proposed HCVSOA significantly improves model subject-independent predictive performance for kinematic handwriting analysis.