Skin cancer is among the most prevalent cancers in people. Most diagnoses are made visually, with clinical screening, histological investigation, biopsy and optional dermoscopic steps. Melanoma identification in the skin has been described as problematic due to the possible issues resulting from inadequate feature selection and the importance of achieving high levels of detection accuracy. To classify the malignant or average level of skin melanoma from MRI images, the optimal Wolf AntLion Neural Network (WALNN) model as a meta-heuristic-based deep learning (DL) approach is used in this paper. Correspondingly, the novel hybrid algorithm, namely Wolf Pack Search and AntLion Optimization Algorithm Optimization, is developed for optimal feature selection to ensure the performance of the Convolutional Neural Network learning classifier. The proposed approach, WALNN, is evaluated with existing techniques such as Decision Tree (DT), Cuckoo Search and Support Vector Machine (CS-SVM), and Convolution Neural Network (CNN) using the ISIC archive skin lesion dataset. As a result, the proposed methodology is carried out with a better outcome in term of sensitivity, specificity, Precision, Recall, and Accuracy and perform the recognition precisely.