Abstract
With the deepening shortage of medical personnel, monitoring individual health status becomes increasingly critical. This study introduces a self-diagnosis and personalized consultation chatbot system, primarily tailored for skin diseases. The system comprises two stages: image classification and chatbot interaction. Initially, skin lesions are classified from images using a model. Subsequently, personal information and the classified skin lesions are input into a chatbot designed to recognize characteristic symptoms of skin diseases. We employed the LLaMa-2 model as our chatbot, fine-tuned it through LangChain, thereby enhancing its ability to deliver personalized information to users. Experimentally, diverse models were employed for the classification of skin lesions, with each model being trained using the HAM1 0000 dataset. Furthermore, our fine-tuned chatbot model exhibits a higher degree of personalization and accuracy in generating sentences compared to the original version.