<p>In recent years, the number of dermatological disease cases has been growing rapidly. The symptoms of these diseases vary from one individual to another, depending on their visual aspects. The skin disease of one type seen on certain body parts may vary in other body parts and several types of disease in one part are similar in other parts of the body. In the medical field, skin diseases, such as skin cancers, are categorized into distinct groups. Hence, it is essential to detect skin diseases in their early stages to prevent them from spreading to other parts of the body and also to avoid the associated risk factors. In this study, an efficient framework for dermatology disease detection is developed. Initially, the required input dermatological images are gathered from benchmark websites. From the gathered images, the necessary features are extracted. The retrieved features comprise texture, morphological, and color features. These three features are concatenated and then the optimal features are chosen from them via an improved chef-based optimization algorithm (ICBOA). Further, the optimally selected features are utilized for detecting the dermatology disease. A hybrid machine learning (H-ML) framework is then developed, which is implemented by combining a weighted probabilistic neural network (PNN) with a capsule neural network (Caps Net) to detect the dermatology disease. The detection results from the implemented dermatology disease model using deep learning approaches are analyzed with other existing detection techniques and optimization algorithms.</p>