Cloud-deployable health data mining using secured framework for Clinical decision support system

Reliable, scalable and secured framework for predicting Heart diseases by mining big data is designed. Components of Apache Hadoop are used for processing of big data used for prediction. For increasing the performance, scalability, and reliability Hadoop clusters are deployed on Google Cloud Storage. Mapreduce based Classification via clustering method is proposed for efficient classification of instances using reduced attributes. Mapreduce based C 4.5 decision tree algorithm is improved and implemented to classify the instances. Datasets are analyzed on WEKA (Waikato Environment for Knowledge Analysis) and Hadoop. Classification via clustering method performs classification with 98.5% accuracy on WEKA with reduced attributes. On Mapreduce paradigm using this approach execution time is improved. With clustered instances 49 nodes of decision tree are reduced to 32 and execution time of Mapreduce program is reduced from 113 seconds to 84 seconds. Mapreduce based decision trees present classification of instances more accurately as compared to WEKA based decision trees.