Melanoma is a type of skin cancer that affects melanocytes, which are the cells responsible for the production of melanin. This neoplasm has the highest mortality rate of all skin cancers. During recent years there has been an increase in new cases, for example, in Colombia the figures given by the High Cost Account placed it as the eighth most frequent tumor in 2021 according to the number of registered diagnoses, and mortality increased by 30/% compared to the previous year. Nowadays there are a large number of methods and techniques to identify it in its early stages, one of them being the ABCD rule, which through the physical characteristics of the mole can determine the suspicion that it has cancerous cells, specifically. : If the mole is asymmetrical (A), has irregular edges (B), more than one or two colors (C), and has a diameter greater than 6 mm (D) it has a high probability of being melanoma, this rule has gained a lot clinical acceptance for the identification of this disease. Based on this, the objective of this master's work was the adaptation of models based on deep learning for the automatic estimation of characteristics that can classify moles as benign or malignant, with validation in images obtained from the /textit{ database International Skin Imaging Collaboration (ISIC) Challenge Dataset}. For this, classic image processing techniques were used to calculate the ABCD characteristics of the entire database and then the training task was systematically assessed using: only the characteristics, only the images, and the combination of both sources of information (features and images), with this final approach, was obtained that both features and images performed better performance metrics in training and testing sets.
