The pervasive use of Android devices has resulted in a substantial increase in cyber threats, notably Android malware, which threatens user data privacy and security. Traditional detection methods that rely on static code or behavioral analysis have become less effective as malware evolves with sophisticated and polymorphic features. This research introduces a novel method for improving the detection and classification of Android malware using bio-inspired optimization algorithms. We introduce the Parrot Optimizer (PO) and its hybrid combination with Particle Swarm Optimization (POPSO) to enhance the overall detection accuracy and feature selection. Using POPSO and GWO methods, we evaluated the performance of a variety of classifiers, such as Decision Tree, Gradient Boosting, HistGradientBoosting, Random Forest, and XGBoost, across a range of population sizes and iterations. The PO PSO approach significantly improves detection capabilities, as evidenced by our experiments. Specific classifiers achieve up to 99% accuracy, while the average accuracy improvement is 5-10%. The significance of exhaustive feature selection, robust machine-learning models, and large datasets in developing effective malware detection systems is underscored by these results.