Design of a skin cancer diagnosing web application based on convolutional neural network model and chatterbot application programming interface

Abstract

Skin cancer has become a great concern for people's wellness. With the popularization of machine learning, a considerable amount of data about skin cancer has been created. However, applications on the market featuring skin cancer diagnosis have barely utilized the data. In this paper, we have designed a web application to diagnose skin cancer with the CNN model and Chatterbot API. First, the application allows the user to upload an image of the user's skin. Next, a CNN model is trained with a huge amount of pre-taken images to make predictions about whether the skin is affected by skin cancer, and if the answer is yes, which kind of skin cancer the uploaded image can be classified. Last, a chatbot using the Chatterbot API is trained with hundreds of answers and questions asked and answered on the internet to interact with and give feedback to the user based on the information provided by the CNN model. The application has achieved significant performance in making classifications and has acquired the ability to interact with users. The CNN model has reached an accuracy of 0.95 in making classifications, and the chatbot can answer more than 100 questions about skin cancer. We have also done a great job on connecting the backend based on the CNN model as well as the Chatterbot API and the frontend based on the VUE Javascript framework.