The analysis of medical data plays a critical role in improving diagnostic accuracy, refining research methodologies, and informing decisions regarding the allocation of medical resources, particularly for critical diseases. Artificial intelligence (AI) provides essential tools for analyzing such data to generate reliable predictions. This study proposes a predictive framework for cardiovascular disease that utilizes key risk factors through a hybrid model combining an Improved Particle Swarm Optimization Algorithm with Mutation Criteria (MPSO) and a Recurrent Neural Network-Long Short-Term Memory (RNN-LSTM) architecture. The model's performance was evaluated on two datasets: one from the University of California Irvine (UCI) Machine Learning Repository and another comprising real-world data collected from Baghdad Medical City Hospital and Ibn al-Bitar Hospital. The proposed framework achieved high predictive accuracy, with Data1 yielding an accuracy of 98.36%, precision of 98.48%, sensitivity of 98.48%, and specificity of 98.21%. Data2 demonstrated an accuracy of 98.75%, precision of 100%, sensitivity of 94.12%, and specificity of 100%. These results indicate that the model generalizes effectively across datasets and outperforms state-of-the-art methods in predicting cardiovascular disease, as evidenced by robust performance metrics.

This paper is devoted to the design of a trajectory-following control for a differentiation nonholonomic wheeled mobile robot. It suggests a kinematic nonlinear controller steer a National Instrument mobile robot. The suggested trajectory-following control structure includes two parts; the first part is a nonlinear feedback acceleration control equation based on back-stepping control that controls the mobile robot to follow the predetermined suitable path; the second part is an optimization algorithm, that is performed depending on the Crossoved Firefly algorithm (CFA) to tune the parameters of the controller to obtain the optimum trajectory. The simulation is achieved based on MATLAB R2017b and the results present that the kinematic nonlinear controller with CFA is more effective and robust than the original firefly learning algorithm; this is shown by the minimized tracking-following error to equal or less than (0.8 cm) and getting smoothness of the linear velocity less than (0.1 m/sec), and all trajectory- following results with predetermined suitable are taken into account. Stability analysis of the suggested controller is proven using the Lyapunov method.