Central nervous system (CNS) diseases have witnessed an alarming rise globally, affecting millions of individuals and imposing significant healthcare expenses. Throughout history, indigenous medicinal plants have played a vital role in addressing various ailments, including CNS disorders. In this research, we propose a cutting-edge machine learning approach to accurately predict the effectiveness of medicinal plants in treating CNS diseases. Leveraging the VNPlant200 dataset comprising plant images and associated metadata, we train a convolutional neural network (CNN) to extract profound features. To amplify the discriminative power of these features, we employ matrix-based discriminant analysis, thereby augmenting our model's predictive capabilities. Furthermore, we integrate an ensemble technique that combines multiple classifiers, resulting in improved accuracy with a remarkable rate of 100%. Additionally, we have developed a user-friendly mobile application empowering individuals to identify and classify medicinal plants based on their potential for treating CNS diseases. This innovative work holds great promise in efficiently and cost-effectively identifying and harnessing the ability of conventionally used medicinal herbs to treat CNS disorders.