This invention describes a method using novel microRNA (miRNA) biomarkers, a novel analytical approach, and a novel machine learning based classifier to determine whether a melanocytic lesion on a patient’s skin is a benign nevus or a melanoma. Sample collection can be performed either before biopsy (by collecting cells with adhesive tape) or after biopsy (by fixing the skin sample and microdissection for melanoma cells). In both cases, RNA is then isolated from the sample and a RT-qPCR assay is used to measure the expression levels of a suite of novel miRNA biomarkers, followed by processing using a trained classifier to generate a risk assessment. This process has been validated in a pilot study, and it is currently being tested in a large patient cohort. Unlike previous melanoma diagnosis approaches, this method uses a previously unreported set of miRNAs whose signatures are robust and have high classification potential for distinguishing between nevi and melanoma, rather than between only primary and metastatic melanoma. Furthermore, collection of cells using adhesive tape is lesion-specific and non-invasive, allowing the test to be used successively on the same lesion or patient, and the miRNA from the sample are much more stable than mRNA collected in the same way. Finally, risk level analysis is performed by a newly developed machine learning classifier.