A cloud-based data mining framework for improved clinical diagnosis through parallel classification

摘要

Healthcare Organizations have been dealing with rapidly growing Electronic Health Records (EHRs) and digital images. Maintaining high volumes of medical data leads to scalability issue. Cloud computing provides scalable resources on demand which includes computing and storage as a service. In this paper, the authors propose a model that would enable mining meaningful medical information from a community cloud that connects multiple health organizations. These health organizations work for a common purpose and store health records in cloud. Storing EHRs in cloud makes data maintenance easier. In order to mine cloud-based medical data, classification - a data mining technique is applied on EHRs to diagnose chronic diseases based on patients' symptoms. Classification algorithms like Decision Tress Induction, k-NN classifier and Naive Bayesian classifier are run in parallel across multiple processors (Virtual machines) in cloud. It is believed that Ensemble methods like Bagging or Boosting can be applied to improve accuracy of each classifier in predicting clinical outcomes.