Publication

Abstract : Disease prediction is one of the leading research areas in the field of data mining. There are many algorithms like Latent Dirichlet Allocation, K-Means, association mining and so on, but still, there are rooms for improvement. We propose an optimized algorithm to predict the disease that is yet to affect a person based on symptoms. We offer a new algorithm called SBD-LDA (Set Based Disease prediction Latent Dirichlet Allocation), which works in 4 stages, in the initial phase, the proposed algorithm identifies synonyms and groups similar words into unique topics. The next step is to extract disease sets that are matching with the unique topics, the third stage is to assign a weighted value for each disease set, and finally, a graph visualization is created to show the predicted disease that is yet to attach the person with symptoms. Dataset from MedGen (NCBI database) is used for our research, and the experimental analysis proves that our algorithm provides enhanced accuracy over existing algorithms.