A Similarity-Based Disease Diagnosis System for Medical Big Data

The disease diagnosis service has occupied an important position at both the societal and individual levels and has sparked a large research interest within the medical field. In this paper, a similarity-based system for disease diagnosis is developed using big data. Firstly, disease diagnosis data within these images are captured and used to populate a database, after which the big data analysis platform, as a middleware, is used for the storage and Latent Dirichlet Allocation (LDA)-based approach is implemented to find the latent health status. Then we employed the Formal Concept Analysis (FCA) model for similarity query and disease diagnosis for users. Finally the bottom hardware is composed of a distributed computing environment which can provide computation and storage service. Our method improved disease data classification and showed that patient survival and mortality can be derived from the similarity-based system for future decision-making and planning. Authentic datasets from medical databases are used to evaluate our method comparing with fuzzy k-nearest neighbor (FKNN) and intuitionistic fuzzy soft sets (IFSSs). The average accuracy of our approach obtained from 20 datasets is about 11.5% and 7.4% higher respectively. Moreover, ours is about 22% and 118% faster than the FKNN and IFSS methods, which shows the robustness of our proposal.