Skin cancer is a significant global health concern, with rising incidence rates reported in recent decades. Skin cancer falls into two basic categories: non-melanoma (benign) and melanoma (malignant). Merkel cell carcinoma and melanoma are more severe cancers, although basal cell carcinoma and squamous cell carcinoma are the most prevalent non-melanoma kinds. Advanced imaging techniques are necessary for effective therapy because early identification and categorization are critical. Enhancing algorithm accuracy in skin cancer diagnosis is mostly dependent on pre-processing procedures. Following the crucial stages of image acquisition, cleaning, enhancement, and segmentation, feature extraction is employed to extract the essential characteristics of the lesion. Data augmentation enhances the generalization of the model by including noise, rotation, scaling, and other techniques in the training dataset. The study considers skin cancer classification using the Modified EfficientNetB7 Model. EfficientNetB7 is one of the largest models in the EfficientNet series, designed for tasks that demand high accuracy and can benefit from a deeper and wider neural network. It is particularly useful for image classification tasks on large datasets [12]. While comparing the EfficientNet-B7 with other models(ARDT-DenseNet, VGG19, Stacked model, Ensembling, Inception-v3) we are getting accuracy of 96.52%,precision of 96.44%,recall value of 96.82% and F1 score of 96.3%.