Personalized Skin Disease Consultant and Care Recommendation Using Lifestyle-Based Analysis

Abstract: In rural and semi-urban locations, many people have skin problems and diseases and it is difficult for them to see a dermatologist. Early recognition of these skin conditions means the patient can take preventative measures, and go to the doctor when required. This research discusses a new web-based system that will perform preliminary evaluations of skin conditions by using deep learning algorithms. A MobileNet based Convolutional Neural Network (CNN) will classify the various skin conditions, based on images provided by the user. MobileNet was selected for use in this web- based system due to it being a lightweight architecture so that only minimal processing power will be required on mobile devices and web browsers. The system also provides basic skin care suggestions and precautionary information so that users can see what their possible next steps for care may be. The model has been trained using and evaluated using a dataset containing multiple types of skin problems and diseases. Experimental evaluation of the model yielded a validation accuracy rate of 92.4%. The results demonstrate that lightweight deep learning models can be reliable and accessible for preliminary skin disease screening applications. Such a system can help provide early awareness of skin issues and encourage users to seek medical attention from a dermatologist if needed.

Keywords: Deep Learning, MobileNet, Convolutional Neural Network (CNN), Web-Based Healthcare System, Dermatological Screening, and Skin Disease Identification.