Malignant melanoma stands as a formidable manifestation of skin cancer, posing a significant threat to human health. Contemporary dermatology acknowledges the paramount significance of timely detection in mitigating the aggregate mortality rate and guaranteeing that individuals are subjected to minimally intrusive therapeutic interventions. The escalating popularity of computer-aided diagnostic (CAD) mechanisms utilized for the timely identification of skin lesions is becoming increasingly apparent. These systems encompass a multitude of sequential stages, each necessitating a judicious selection that aligns with the inherent attributes of the digital images, thereby culminating in a dependable diagnostic outcome. The successful advancement of automated diagnosis for life-threatening lesions like melanoma necessitates the conquering of various challenges encompassing acquisition, pre-processing, the process of segmentation extraction of features and selection, and ultimately categorization of dermoscopic images. In the suggested method, Random Forest and Resnet-50 are combined. The experiment improved precision by 9-10%.